Empirical Musicology: Aims, Methods, Prospects 019516749X, 9780195167498, 9781423784692

Table of contents :
00000___74488b41761a01c0a45b2cd58c8f4873......Page 1
00001___71afb0628cf78bdc2e8ee794b473a0c2......Page 2
00002___cf0def8f651a1f6b01166f6ebed64698......Page 3
00003___71c0e836df67d0723853161609ffbc88......Page 4
00004___ecf8d8ddf517ae3d5568ff206d92a305......Page 5
00005___b845c3ffacb8f1914acd83cd653f1edf......Page 6
00006___703cfd92677084a6823201124564603f......Page 7
00007___b772bd8f7fb2fbcfa4730d0e0334c8c1......Page 8
00008___e347e7dd59630cc24aa1218117e79890......Page 9
00009___f2acf7903680a8108785e29a833c5e65......Page 10
00010___39c1e6e1779225729df88afa258401f0......Page 11
00011___f6f37c4a68370ade7333e378a78fab6c......Page 12
00012___ed136fd4848198273e10f101de5b1cd2......Page 13
00013___2a3738f31601537f8953abbecd9d47a9......Page 14
00014___21158a32df4c664b9b58088e4f468266......Page 15
00015___6b5ca8b440ad8ea6287dae4d7f4d7117......Page 16
00016___f97ea39ba9e90328b652c656eaa588e2......Page 17
00017___62761e29b6f8c9de667b7f4fa127d157......Page 18
00018___91fec38a409ab6cdd81898d1336cdefb......Page 19
00019___9432f7c4386a3ad62b487b8173b8d799......Page 20
00020___d89a737132dd51ca445bc91b6838ca33......Page 21
00021___32c5f49652bd6def88592db182572384......Page 22
00022___4bf328428b5d93af5f30063882a2553a......Page 23
00023___a8ace019a70bd4db02b581ea02ac9c20......Page 24
00024___70fa89c29d4daa02a7a1fdf3f4656b6e......Page 25
00025___8ee707c8164a1751783508f16cc7f4cb......Page 26
00026___e2334d0489ff71ccb4ca6509c39d4c60......Page 27
00027___410d942dc744182739e426cc0b087e62......Page 28
00028___2b1d849781b35e4c75287106fb0acdec......Page 29
00029___322bcf909b26d84b542fd79dd9a039df......Page 30
00030___8bef9749ad2c4f5cebb5672f91363579......Page 31
00031___a7cea314684872d2fc0da869e975df88......Page 32
00032___c9a6a2fd854e389d57cc3ee610c2dba0......Page 33
00033___6230e9df58d6c0669d159c16cebeea43......Page 34
00034___f20b85b3c47eeedfc29d7719369f29de......Page 35
00035___0d21c6389053fa29acb55d8ca68e0e21......Page 36
00036___3a13ccc4697a607d9ce16ddc53fed1a9......Page 37
00037___dac4c2ae9706d8d2db283bcb36ca3d92......Page 38
00038___81871617902d6c93fd92bb83b9d4b592......Page 39
00039___41ba1d9f40abc2d08b48c2596f5b19a3......Page 40
00040___5e75fd800b637757f1303f02a9ab5c04......Page 41
00041___05d9790489aed88e22a3568d1bef232c......Page 42
00042___6765ed0d3d247d92ff77a8b51dc8e4d2......Page 43
00043___2279686386b622d7509ae7ff52b2b79a......Page 44
00044___d345af2ab3943afe1b75c55aed4d471d......Page 45
00045___e14f1b4fd293a47fd022163c7378c63d......Page 46
00046___37f14cb3336c9f9aad6f57fb5cf63a5a......Page 47
00047___c79a86dff51f86a0037330b93afb9429......Page 48
00048___327023f8cf2b15e36bfc464f364478f4......Page 49
00049___f7a96bb035428f9d8dc4ef61e8f00299......Page 50
00050___43161a39646705691835e550a2e22e7d......Page 51
00051___ba0ca3f6dc931ffd7f153c016a707c7c......Page 52
00052___283ef928effa4f8e91e3915d7af427b2......Page 53
00053___fe2e985bc5f6645dc6f46a55328f7195......Page 54
00054___4233033403ef85f269b5c6d1530db463......Page 55
00055___37b775aee1e29ba05b5c49e7209e2c6b......Page 56
00056___52cbddbc9b658b8d3e08facace5a9f91......Page 57
00057___6e910b8cbfa7c1b67e99dc3c5d6cc6ac......Page 58
00058___f507f5432ddaed4b7f77c5757d95a695......Page 59
00059___44ad8b55d0045ab52e90449777db58ef......Page 60
00060___cab74ae38d37535ec4fc9e68211789f9......Page 61
00061___ef4735fb745e6eb9d9709df16d698b19......Page 62
00062___dc17fd64e4cf842bc36901d75f2ae5f1......Page 63
00063___8467df0c8addba25ee0789aa34d0f59e......Page 64
00064___1f1a03464b66af5dac0a3fdffa04b400......Page 65
00065___ba41ffd552d48d6170ba28d1e9350e9c......Page 66
00066___723085b5e52a335c4836ad3a5f1db108......Page 67
00067___76e16b98f4ecc5d9eeae6972bebeb9a1......Page 68
00068___ee37308b651e52194b775c7180034cf6......Page 69
00069___f930c70b686d639227d6a3a937bbeccf......Page 70
00070___d24ca1781138fc6c6ef1807ca6e82b97......Page 71
00071___0eb5eefd0671c4b6d9d182bd0315567a......Page 72
00072___956d90aa5d4b298a3e5dba9125c136e0......Page 73
00073___1a8902630b2d5fd1dc5a24b4c57f8570......Page 74
00074___caa926bdf11d185eb7ed7fe9ca8bb25e......Page 75
00075___9f2ed309e5ca063a4f4b4e87117281b0......Page 76
00076___50357b116eca42b7ea89fd9a6b2c9b89......Page 77
00077___1d36416a768378811f45d17d51948cd5......Page 78
00078___b254c136820a43417e69b37142bb520b......Page 79
00079___7ea630a76da469008b02a71fc8dc9ca6......Page 80
00080___2251fd8f75b0e3f9f285baeec1ff3ed7......Page 81
00081___ceac73a66d2a84b2d1ca3ee3163dcfbb......Page 82
00082___b2618aa31c31ea5a3d77cb87b51c5a59......Page 83
00083___c6a0fd8e80906fdb4410cf6ab89bbf69......Page 84
00084___bf8ccb138d2c79a6ec41ba1023d17927......Page 85
00085___e2a69e9dd8461499a6cd672a5d0c80e4......Page 86
00086___059bce73c2f4926ed401379f980a993d......Page 87
00087___ad1e33c552bb6435607eefa870eed7f2......Page 88
00088___a3cdacdb8478ca708519ccd8db290210......Page 89
00089___eea482de256031f2b86e3a60836fc72e......Page 90
00090___45ce7a5eadb1c40d5ca71e5c0cf07cd2......Page 91
00091___3cd1d5bc127a05132640f869ae3a3f0e......Page 92
00092___0fbc71649dd5dcaf0388c0187501e011......Page 93
00093___05f1d4053174ca38944553e9a7945531......Page 94
00094___44916282035e5ef6829c4aabb84b7e18......Page 95
00095___44071848e74fc8e08265cdbc09faf9ee......Page 96
00096___b9000f1f8929e9f633aca5db78fc3260......Page 97
00097___cd3188451974ef51ebd035b81b31367e......Page 98
00098___e10cd00a89131b81e3dc07963c717833......Page 99
00099___c947b0f587a4755a035237d8359db512......Page 100
00100___8d48e501807555b93c6e9d46c351eadb......Page 101
00101___e9ef073ff565bb1d30fd7752cb5b4204......Page 102
00102___b5feef47e03e0e4c22edc34dacd3bc6c......Page 103
00103___7398f3c68c9e0050c1910b6d41d1416b......Page 104
00104___07a21209c78d5650b294ac6fa71a6665......Page 105
00105___c667475a49c0849d991be50a925a627d......Page 106
00106___5e8b07f9bc930cba5ec2f78b3d27e8e3......Page 107
00107___70733398ae79bf4ff912fdb6d2312480......Page 108
00108___811b21ae9c7a12e03ca3f72d48457f5f......Page 109
00109___3b8dbe06f48b8004b610acd941d08d2f......Page 110
00110___6d8e0a75f3f18ffad5206f7e6600a9ef......Page 111
00111___d93dc71f3c437d5ae821964ade6bf053......Page 112
00112___99c9e0ab90b9c70b606db8f6302272f2......Page 113
00113___2d15c2b5ed5ce5ecd72d1e64682470b2......Page 114
00114___7dc6a71932058d4ace33a43c0496ad6f......Page 115
00115___d1c399f5885691c8965156f973c939c3......Page 116
00116___b090b0aeefe1ac53feec2b7afe4a54a1......Page 117
00117___50afe1d6b6ffde89f378b7a82f7ba81a......Page 118
00118___cd3b7fcf4832e16cc26d37f3b0faaf63......Page 119
00119___36e20f182bd3ebacde31b111844fa21b......Page 120
00120___ce8702f695f214c59e18fd942e876cc6......Page 121
00121___246e4890c03e3fcd56b0a894507fc507......Page 122
00122___79a79ae019669743e951549aa027a7f3......Page 123
00123___4b3c2997cbde02e09f40ba6ccd478192......Page 124
00124___5f8c206715910da48cfdba10822a9289......Page 125
00125___b02b25dc44a549dd4afcf9e0872a539b......Page 126
00126___9ae880d357402ea765232b2e2384038e......Page 127
00127___58029eba0467c794becd2b9d2a969107......Page 128
00128___a4880081bb3b6977be380d50776dd675......Page 129
00129___dc923066061c0ee60ad98e51688180f7......Page 130
00130___46c136e42e43ee0881617f30d25bb6c7......Page 131
00131___86ed7d31a7d0c315d70f9fcb40ad73d6......Page 132
00132___83e697b0b6b43862156072be1a53312f......Page 133
00133___87798a01a4c0ace72bfb6a90f9649175......Page 134
00134___49dceba4637e852d1e2c3e7776f9c069......Page 135
00135___0415330d136a6b5f34cf3922dd8ae5a6......Page 136
00136___724b5582615d7cb2e1d92903a1d1479d......Page 137
00137___b39854eccb5777510a40dab3373765f2......Page 138
00138___15a40ad2334bf25d9be62b1290d4c186......Page 139
00139___05b8a9dcf13d4e2353ae23010373dc7c......Page 140
00140___26f261c3498e7c644b679e127971bc86......Page 141
00141___280b80a8d3065cd54915b5e396f338a8......Page 142
00142___2ef47feea003e7d329a383c164ba42be......Page 143
00143___21a26e428046d83852ea14cb3950cdda......Page 144
00144___1ff1eeb3192a831d0534a9c26efcc8b3......Page 145
00145___0adc7448d74428e395604a45efe5d704......Page 146
00146___ff4cb5b464d2bd19e08fa3fe3a7e78b0......Page 147
00147___cabb32a2aab46a1e09c1d98f56c737f4......Page 148
00148___6d2ffd56423d8dfa593b691b67323cea......Page 149
00149___35c7d8887f24541820aa271ac465a1e0......Page 150
00150___25c609fb66a1994e9f90f9ebad79320a......Page 151
00151___7fda38e59622f3eb2770bc923216b104......Page 152
00152___d7864a6d4602b9f0d55d69dbc0ae07a3......Page 153
00153___38242c1593ddf98a4d6ed2ed59e43c64......Page 154
00154___fa46701101aebf8f7d5120442cf6b3ea......Page 155
00155___2fe1622de809eada2b435661b81aaefa......Page 156
00156___1dbe9c6a25fabc1759cc6305ce1db503......Page 157
00157___ac04c32661a93aee7722f4f690cc476f......Page 158
00158___c56043fb8910c5af5f63d07b9e95588e......Page 159
00159___b247427fb7f6b0714d03a49a259aa0be......Page 160
00160___cec206e98d9efbe0206870585282634d......Page 161
00161___ff8abbbbe3ab570e148c3b63b13e0952......Page 162
00162___b8c4549298816d4060a9f18e1218a39e......Page 163
00163___e0ac3ab0124e6b8a2610847668f1eeb9......Page 164
00164___a4dfd0b9d17633b20ade95648fe8cbbc......Page 165
00165___1def48c5ebd283bf0e713bac17878a5f......Page 166
00166___387f73c14b866957bf0c0bf8f7d4acb2......Page 167
00167___4f484c68b1fdfaa7eef98b64004d33a3......Page 168
00168___c834127d224d05b3c464777e8c3d1e26......Page 169
00169___7e1fcbdf0e2c5889b5724930ebd69575......Page 170
00170___192197282fe1a2235173052c35d916e6......Page 171
00171___0681600ed4d929d2c6354fa990066651......Page 172
00172___a98dd9f68db0fffe6763d3031ab408bb......Page 173
00173___75b2765ebecd3f8d03e752ff23a74eb9......Page 174
00174___b8586a91daed4aaf8ad3246cfa7ee526......Page 175
00175___fccce36e427f9d286481dbc4fe691dea......Page 176
00176___fd64a784d20d867cb7608cd311038843......Page 177
00177___0f60e349a18a4fffdf02893c88cbdd89......Page 178
00178___78348777eb76062635a155c0bd5aa9c1......Page 179
00179___5a5a88c53b66a20c05d2009d022aae98......Page 180
00180___fbe509d852b306402e3c30e3d77df2e2......Page 181
00181___054f9b47298f21431d33c52ac70a005c......Page 182
00182___efec75ce104f0318e0638baef94f5588......Page 183
00183___ae4f1cc87296b57746e9f4097ae2290e......Page 184
00184___9c5c7a467b42b76d67d25f325e471d24......Page 185
00185___aeb7588f7eff3b9599929a396ffb4a82......Page 186
00186___73ea628e018aef721e535a543784da9d......Page 187
00187___a844aa8b60c0b7b531d90f7a9704a4f7......Page 188
00188___4880b3a91ba85b9e1d18f7fb23eac9bb......Page 189
00189___8ad698e9a14860f9475fe098ad7eaecc......Page 190
00190___813ec0eb5175bc476388711a4549dfd1......Page 191
00191___074204c42b5ff1e69e1247e828725955......Page 192
00192___14e79d6ed5e89153ed8a0d6a011def2c......Page 193
00193___cc2bef7905964782240e9d104c415207......Page 194
00194___aaade71c139a0e4702d1fe5d41c75dae......Page 195
00195___932bdd774c520b71daffb77debaef61a......Page 196
00196___844eb5185c7fd4c7d11b31c7e0a0cc86......Page 197
00197___4d9b02f08465c4921acf0f5dbd0889d3......Page 198
00198___095925c0a8832f189875c0bb797a3ac5......Page 199
00199___f5afb983aaff56dbdf7de13f6bcdfda2......Page 200
00200___d22e59eebbf25c453297ee1bef7b6675......Page 201
00201___652e66ced3bec0ade7306aa789bc45f8......Page 202
00202___ec5975b5d46ef4b6a2f43953b2874ec3......Page 203
00203___0b9cdb5902a3bfdcaba0f1af1881520f......Page 204
00204___3df11e1e6de0d61265b9c38efaf21ec2......Page 205
00205___e19d25ebbf294808f20a904882b2afd6......Page 206
00206___3560bf9d7977eed5fbbe9bd05ebc3851......Page 207
00207___81ea8f300549dcabb4e3d40f290e225b......Page 208
00208___32257c10e09fbd9bbbff4fbc30b15ded......Page 209
00209___8fba0edca45922f58da3ad1ca6d969c7......Page 210
00210___be001c0c884393095746035c590d3eb4......Page 211
00211___17ddf69618a32865926aba2cc20cd624......Page 212
00212___f72a93ab57d4f9b69c9d477cfff2fbf7......Page 213
00213___2ac3c4f22521ed82559e6d1c9ae95476......Page 214
00214___d464259b8ca072c1ede7a0c1dfb58b65......Page 215
00215___14971a3062077073a3a236582b39f587......Page 216
00216___0f860cdffe6c3e98420e9555f4acef02......Page 217
00217___b338e38ec5b1860921bc0dc9fb8de8c8......Page 218
00218___84924af1fcefc02d18abdd258ff36747......Page 219
00219___8890c09aab11bd1f2abb5739810ae6db......Page 220
00220___9b2185af3f55fd08cc8927d6a7f8345e......Page 221
00221___b8021773fe2ce6aca2eacbd2e815ac82......Page 222
00222___c94de79296fdf68bc979003364ec5139......Page 223
00223___434479421ee68d98c7bea389fa23b1fd......Page 224
00224___cd893a8148fa2549c63e2efdb64ab465......Page 225
00225___34325695324b8c383d65435460263706......Page 226
00226___b78e774f758f5227657aa7e9337b6e89......Page 227
00227___e783373d97ba1cf1a085e3e30f19e471......Page 228
00228___5e325996fc1380f0db7423e27d25dad1......Page 229
00229___954f30e2b04c5de1767faf8293c43210......Page 230
00230___7c848ad03327498ae32d6903fb2f0ac9......Page 231
00231___1a8d49ab960dbb2684f7dcf50ba4a6fa......Page 232
00232___fbc42eb7684090bb1f4ab15d3fe3dbfb......Page 233
00233___ac3d7afc7189c491b7b559f23b0b7c78......Page 234
00234___8d9fa838add94f9a6f42770739e271bb......Page 235
00235___0123816d52368d1a45debdd5c37bb532......Page 236
00236___ac5a93bf70060e61b595c5fffe73d4ef......Page 237
00237___97c24d79a459bb75a2233c7f5dc3e62a......Page 238

Citation preview

Empirical Musicology: Aims, Methods, Prospects

Eric Clarke Nicholas Cook, Editors

OXFORD UNIVERSITY PRESS

EMPIRICAL MUSICOLOGY

This page intentionally left blank

Empirical Musicology Aims, Methods, Prospects EDITED BY

Eric Clarke and Nicholas Cook

1 2004

1 Oxford New York

Auckland Bangkok Buenos Aires Cape Town Chennai Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi São Paulo Shanghai Taipei Tokyo Toronto

Copyright © 2004 by Oxford University Press, Inc. Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 www.oup.com Oxford is a registered trademark of Oxford University Press All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Two figures from F. Lerdahl and R. Jackendoff, A Generative Theory of Tonal Music (Cambridge, Mass.: MIT Press, 1983), pp. 259, 260, are reprinted with permission from MIT Press. Figure 8 (Pitch/contour schemata recognised in early Mozart by Robert O. Gjerdingen’s ART pour l’art neural network, p. 360) from R. Gjerdingen, “Categorization of musical patterns by self-organizing neuronlike networks,” Music Perception 7 (1990): 339–369, is reprinted with permission from Blackwell Publishing Ltd. The segmentation graph of Berg, “Warm die Lufte,” bars 19–21 (p. 199), from J. Doerksen, “Set-class salience and Forte’s theory of genera,” Music Analysis 17 (1998): 195–205, is reprinted with permission from Blackwell Publishing Ltd. Quotations from pp. 49–54 and figure 18 (Harmonic analysis of “Yankee Doodle,” p. 63) from D. Temperley, “An algorithm for harmonic analysis,” Music Perception 15 (1997): 31–68, are reprinted with permission from the University of California Press. Library of Congress Cataloging-in-Publication Data Empirical musicology : aims, methods, prospects / edited by Eric Clarke and Nicholas Cook. p. cm. Includes bibliographical references. ISBN 0-19-516749-X (hardcover); ISBN 0-19-516750-3 (pbk.) 1. Musicology. 2. Empiricism. I. Clarke, Eric F. II. Cook, Nicholas, 1950– ML3797.1.E47 2004 780'.72—dc22 2004012210

9 8 7 6 5 4 3 2 1 Printed in the United States of America on acid-free paper

CONTENTS

CONTRIBUTORS

vii

1 Introduction: What Is Empirical Musicology? 3 Nicholas Cook and Eric Clarke

2 Documenting the Musical Event: Observation, Participation, Representation 15 Jonathan P. J. Stock

3 Musical Practice and Social Structure: A Toolkit 35 Tia DeNora

4 Music as Social Behavior 57 Jane W. Davidson

5 Empirical Methods in the Study of Performance 77 Eric Clarke

6 Computational and Comparative Musicology 103 Nicholas Cook

7 Modeling Musical Structure 127 Anthony Pople

8 Analyzing Musical Sound 157 Stephen McAdams, Philippe Depalle, and Eric Clarke

9 Data Collection, Experimental Design, and Statistics in Musical Research 197 W. Luke Windsor Index

223

This page intentionally left blank

CONTRIBUTORS

Eric Clarke, Professor of Music at the University of Sheffield, has published on the psychology of performance, the study of rhythm, and musical meaning. Recent publications have focused on motion in music, artificial modeling of expressive performance, meaning in pop music by Pulp, Frank Zappa, and P. J. Harvey, and the relationship between music, psychology, and cultural studies. He was Chair of the Society for Research in Psychology of Music and Music Education from 1994 to 2000 and is on the editorial boards of Psychology of Music, Music Perception, Musicae Scientiae, and Music Analysis. Nicholas Cook, Professor of Music at Royal Holloway, University of London, has published journal articles on a wide range of musical topics from aesthetics and analysis to psychology and pop. A Fellow of the British Academy and former Editor of the Journal of the Royal Musical Association, his books include A Guide to Musical Analysis; Music, Imagination, and Culture; Beethoven: Symphony No. 9; Analysis through Composition; Analysing Musical Multimedia; and Music: A Very Short Introduction. He is Director of the AHRB Research Centre for the History and Analysis of Musical Recordings (CHARM). Jane W. Davidson, Reader in Music at the University of Sheffield, completed M.A. and Ph.D. degrees in Music at City University while working as an opera singer. After jobs at Keele University and City University, she took up her current post at Sheffield in 1995, where she lectures and researches on psychological issues related to performance. She has published very widely on this and other topics, and was Editor of the journal Psychology of Music from 1997 to 2001. She maintains an active career in performance and direction, working on projects from opera to dance, and completed an M.A. in Contemporary Dance Choreography in 2002. Tia DeNora teaches Sociology at the University of Exeter, and has special interests in music sociology. Her books are: Beethoven and the Construction of Genius (1995); Music and Everyday Life (2000); and After Adorno: Rethinking Music Sociology (2003). Philippe Depalle received a Ph.D. degree in Acoustics from the Université du Maine, Le Mans. He was an assistant professor from 1985 to 1988 at the École Supérieure d’Électricité, and a researcher in the Analysis/Synthesis team at the Institut de Recherche et Coordination Acoustique/Musique (IRCAM) from 1984 to 1997. He was visiting professor at the Université de Montréal from 1997 to 1999, and since 1999 he has been Associate Professor in the Faculty of Music at McGill University, Montréal, where he chaired the Music Technology Area from 1999 to 2002. His research interests are related to the modeling, synthesis, analysis, and processing of sound signals for musical applications.

vii

viii

Contributors

Stephen McAdams received training in music theory and composition, electronic music, computer music, experimental psychology, neurosciences, and hearing and speech sciences. He received his Ph.D. in Hearing and Speech Sciences in 1984 from Stanford University. He has worked since 1981 at the Institut de Recherche et Coordination Acoustique /Musique (IRCAM) where he founded the Music Perception and Cognition team in 1984. He is currently Research Director in the Centre National de Recherche Scientifique (CNRS) and will direct a new laboratory for research in Auditory Perception and Cognition in the Cognitive Studies Department of the Ecole Normale Supérieure in Paris. His research interests include auditory scene analysis, timbre perception, ecological acoustics, and music cognition. Anthony Pople was a Professor of Music at the universities of Lancaster, Southampton and Nottingham. He was Editor of Music Analysis from 1995 to 1999 and coauthored the extensive entry “Analysis” in the 2001 Edition of The New Grove. His many publications include books on Berg, Messiaen, Scriabin, and Stravinsky, together with studies of works by Vaughan Williams and Tippett. His Tonalities project uses sophisticated but user-friendly computer software to engage with the analysis of music in widely varying styles beyond common-practice tonality. He died in 2003. Jonathan P. J. Stock is Reader in music at the University of Sheffield. He is author of three books: Musical Creativity in Twentieth-Century China, on the key folk musician Abing (1996); World Sound Matters (1996)—a three-volume anthology of transcriptions, recordings, and notes for high school teachers and pupils; and Huju: Traditional Opera in Modern Shanghai (2002). Recently published articles discuss the interplay of speech tone and melody in Beijing opera, ethnomusicology and “new musicology,” and timbral views of formal structure in Mozart’s piano concertos. He also edits book reviews for The World of Music. W. Luke Windsor’s research has focused on the perception and production of musical rhythm and meter, expressive and cooperative timing in piano performance, and the analysis and perception of electroacoustic music. He is a senior lecturer in the School of Music at the University of Leeds, where he has worked since 1998. Prior to this he worked as a researcher in the UK and Netherlands, was involved in course development at the University of Sheffield. He is the coeditor, with Peter Desain, of Rhythm Perception and Production (2000).

EMPIRICAL MUSICOLOGY

This page intentionally left blank

CHAPTER 1

Introduction: What Is Empirical Musicology? Nicholas Cook and Eric Clarke

I believe there’s a real world out there, because not all of my fantasies work.

The words are those of the composer and music theorist Benjamin Boretz (1977: 242), and they articulate a level at which it is hard to envisage work in musicology or theory that is not empirical. Any archival musicologist, after all, knows that facts can be very hard indeed—though, as we shall see, it would be more correct to say that facts are a matter of interpretation and that it is the data that are hard. In the same way, different analysts’ Schenkerian interpretations of a given passage may well differ (and Schenkerian analysis is supported, or at least surrounded, by a discourse that is largely speculative if not at times metaphysical), but they are closely regulated by the score on which they are based; indeed the trial-and-error process by which music-analytical interpretations develop, with observation leading to interpretation and interpretation in turn guiding observation, is a model of close, empirically regulated reading. Theorists and composers have both on occasion invoked the language of experimentation, too; for example, Marion Guck (1994: 62) has described her analyses as “(thought) experiments,” but the best known of such invocations is Milton Babbitt’s (1972a: 148) claim that “every musical composition justifiably may be regarded as an experiment, the embodiment of hypotheses as to certain specific conditions of musical coherence.” In short, there is no useful distinction to be drawn between empirical and nonempirical musicology, because there can be no such thing as a truly non-empirical musicology; what is at issue is the extent to which musicological discourse is grounded on empirical observation, and conversely the extent to which observation is regulated by discourse. The idea of regulation is essential in this context. Michel Foucault (1970) has illustrated this point through reference to the comparative illustrations of human and bird skeletons published in 1555 by Pierre Belon: as Foucault says, these illustrations look like the products of nineteenth-century comparative anatomy, but the resemblance is little more than chance, because the interpretational grids of sixteenth-century and of nineteenth-century thought are so different.1 In other words, what we generally think of as empirically-based knowledge — as science—depends not only on observation but also on the incorporation of observation 3

4

EMPIRICAL MUSICOLOGY

within patterns of investigation involving generalization and explanation. (That is what turns data into facts.) It also depends on the more fundamental criterion of replication: if an observation is to be regarded as trustworthy, it should be possible to make it on different occasions, and it should be possible for different people to make it. The issue, then, is whether musicology fulfils these conditions — whether, in short, its interface with Boretz’s “real world out there” is as well managed and understood as it might be. Musicologists are certainly aware of the distinction between data and facts (see, e.g., Dahlhaus 1983: 34). Researchers with a background in the hard or social sciences, however, might well question whether most musicologists are sufficiently aware of the methodological consequences of this distinction. Like most humanities scholars, musicologists are prone to build interpretations on very small data sets or even on single instances, and the less the evidence that has survived from the past, the stronger this tendency will be. In the study of medieval music, for instance, so little documentation has survived that what does exist often lacks a secure context, and under such conditions it becomes impossible to avoid circular argument: if your starting point is that there are hidden meanings in fourteenth-century motets, then you are bound to deduce that there were sophisticated contemporary audiences capable of appreciating them, and this then becomes evidence for the hidden meanings (Leech-Wilkinson 2000). Without sufficient evidence to prove or disprove the hypothesis, it is simply not possible to cut through the circle; the problem is endemic. It follows that, as David Huron (1999) has pointed out, the issue is not one of good or bad methodology, but of what is viable in data-poor as against data-rich fields. In most (though not all) of the physical and social sciences, it is possible through systematic programs of observation to acquire large bodies of data, which may then be manipulated statistically and subjected to measures of statistical significance. But in fields like medieval music, this is simply not possible, and so arguments based on statistically insignificant samples or single instances are inevitable; a further result, arguably, is that scholars become wedded to their interpretive hypotheses, since there is rarely the evidence to conclusively overthrow them, resulting in a degree of conservatism that can easily turn into dogmatism.2 That is the price that has to be paid for working in data-poor fields. While this is no argument for abandoning such fields, there would be grounds for legitimate criticism if musicologists working in data-rich fields did not take full advantage of the methods available under such conditions, instead restricting themselves to traditional “humanities” approaches developed for data-poor fields — and one of the messages of this book is that musicology is or could be, in many instances, a significantly “data richer” field than we generally give it credit for. (More bluntly, there may be many musicological certainties that would not survive a systematic engagement with the available data.) The same applies to a second characteristic of most work in musicology, which is its retrospective nature. One of the obvious determinants of historical method is that you can’t run history again under different conditions and see how it turns out. (What would the history of nineteenth-century music have been like if Mozart had died in 1845, at the age of 89?) Once again, there is no point complaining about this; it is simply how history is. You could reasonably complain, however, if there were areas of musicology in which prospective work —

Introduction

crudely, making predictions and then testing them—would be possible, but was not carried out, perhaps as a result of the discipline’s predominantly historical selfimage. Empirical musicology, to summarize, can be thought of as musicology that embodies a principled awareness of both the potential to engage with large bodies of relevant data, and the appropriate methods for achieving this; adopting this term does not deny the self-evidently empirical dimension of all musicology, but draws attention to the potential of a range of empirical approaches to music that is, as yet, not widely disseminated within the discipline. And just as it is not a matter of empirical versus non-empirical, so we do not wish to draw an either / or distinction between the objective and the subjective. In order to illustrate this point we may return to Guck and her “(thought) experiments.” She coins this term with specific reference to Hans David’s description of the C
in bar 53 of the second movement from Mozart’s G minor Symphony K 550 as “unexpected” (a description that Babbitt had in the 1960s characterized as an “incorrigible personal statement”),3 and goes on to outline a way of thinking about the movement that explains this “unexpected” quality: she likens the C
to an “indomitable immigrant” (1994: 72), at its first appearance conspicuously foreign to the tonal environment of the movement, but eventually assimilated within it and even ultimately serving to transform it (“C
has succeeded to the leadership of its community” [1994: 70]). In short, she describes a way in which she can hear the music, and invites her reader to share her experience. In what sense can this be properly called an experiment, “thought” or otherwise? There is no null hypothesis,4 no control or randomization of potentially extraneous variables, no control group. To say that, however, is not to say — as a perhaps too casual reading of Babbitt might imply—that it is an exercise in uncontrolled, purely subjective speculation. If Guck’s frankly fictive account of the immigrant C
articulates a way of hearing the music that other people can share, then it can be regarded as a discovery procedure resulting in a replication of experience, and hence in a measure of intersubjective agreement. And indeed resort to measures which are replicable but not necessarily definable in objective terms is quite normal in musicology. An example is the coding of folk songs employed in Alan Lomax’s Cantometrics project, which involved a large number of researchers scoring recorded songs for such qualities as nasality: Lomax explained that we don’t know how to define nasality in objective or productional terms—but what matters, he said, is that in practice there is “good consensus on the presence of great nasality or its relative absence” (1968: 72). Guck’s and Lomax’s arguments, however, would have cut little ice within the culture of objectivity that characterized much postwar musicology and theory (and in which some of the origins of empirical musicology are to be found). Rather like the compositions at the contemporaneous Darmstadt Ferienkürse für neue Musik, such work reflected a distrust of conventional approaches and even terminologies; the traditional language of musicology seemed hopelessly compromised by latent subjectivity, and so it seemed necessary to reduce analytical statements to objective propositions or, if this couldn’t be done, to abjure them. Arthur Komar (1971: 11) referred to “designing a set of rigorous terms for music” as “the serious but unfulfilled goal of current music theory,” but the definitive statement, dating from 10 years earlier, was once again Babbitt’s (“there is but one kind of language, one kind

5

6

EMPIRICAL MUSICOLOGY

of method for the verbal formulation of ‘concepts’ and the verbal analysis of such formulations: ‘scientific’ language and ‘scientific’ method” [Babbitt 1972c: 3, but originally published in 1961]). And when translated into practice, the results included attempts to implement existing analytical approaches as computer programs (Kassler 1967), or to express new ones in the forms of symbolic logic (Boretz 1970) or machine-readable algorithms, as in the case Allen Forte’s (1973) pitch class (pc) set theory. They also included some picturesque attempts to extract rigorous content from such “incorrigible” ordinary-language statements as the claim in Cobbett’s Cyclopedic Survey of Chamber Music that “the spirit of nationalism is felt in all of the best chamber music”: Fred Hofstetter, a pioneer of humanities computing, reduced this to the more straightforward claim that “composers differ from one another as a function of their nationalities” (Hofstetter 1979: 105), selected a representative sample of chamber music by French, German, Czechoslovakian, and Russian composers, defined a stylistic measure (the relative frequency of different intervals), and did the sums. His conclusion, perhaps unsurprisingly, was that there are indeed differences between national styles, and that the most distinctive style is the Russian. The conclusion, of course, was not the important thing. The point of Hofstetter’s project was to demonstrate that informal statements about music could be reduced to formal propositions, in which form they could be subjected to rigorous testing. And it is this ideologically motivated idea of reduction that seems so foreign from the vantage point of the twenty-first century, when this kind of unreflective positivism is no longer widespread, even (one might almost say “particularly”) in the hard sciences. The “nothing but” kind of objectivity embodied in Hofstetter’s project is evident enough, but subtler forms of the same thinking can be more insidious: as Gerald Balzano (1987) has convincingly argued, a chronic problem in music psychology has been the tendency to understand perception as an internalization of objective (e.g., acoustic) categories and structures. Postwar reductionism, then, has left a legacy that has not been entirely shaken off, but it is not the focus of “empirical musicology” as defined in this book: the culture of objectivity in the 1960s and 70s reflected an epistemological world view formed by the ideal of scientific progress, a stance that might be described not so much as “empirical” as “empiricist.” The orientation of this book, by contrast, is intended as an essentially pragmatic one, in which reduction is seen—in Huron’s (1999) words — as “a potentially useful strategy for discovery rather than a belief about how the world is.” In some ways the positivist approaches of the postwar period were more a matter of appearances than of substance; even at the time, Forte’s pc set theory was criticized on the grounds that, for all its apparent objectivity, it was based on analytical decisions about how to divide the music up into segments for which there were no properly defined criteria. But there was a quite different problem that early examples of apparently objective analysis tended to present, which is what one was meant to do with them, what their value was. An appropriate example, since it achieved considerable exposure at the time, was Matt Hughes’s (1977) “quantitative analysis” of Schubert’s Moment Musical in C major, Op. 94 no. 1. This was certainly objective in the sense that a suitably programmed computer could have carried out the analysis without human intervention: it began with a straightforward note count, not in the serial sense but as a simple computation of how many Cs, C s, Ds, and so on there

Introduction

were (that is, it is pitch classes that were counted, and the values were weighted by durations). This provided a measure of the tonal “orientation” of the piece (which turned out to be directed toward G rather than the tonic, C); the values for each pitch class were then mapped onto the cycle of fifths, and analysis of the resulting pattern of peaks yielded a measure of the piece’s tonal “complexity.” What is of concern here, however, is not the details of the method, but the context in which it was presented, and what came of it. The reason for its exposure is that it formed part of a “symposium” on Op. 94 no. 1 published in a widely disseminated collection edited by Maury Yeston; the other elements of the symposium were a “compositional analysis” by Lawrence Moss and two Schenkerian analyses by Carl Schachter and John Rothgeb. The two Schenkerians engaged with one another, but otherwise — as frequently happens at symposia—the contributors talked (or rather wrote) past one another, without any form of mutual communication being opened up. And since then, Hughes’s approach has disappeared more or less without trace. All this amounts to saying that there was little context within which to understand Hughes’s analysis, and that context is essential to value. This can be made clear through a comparison with current work taking place on the border between music theory and cognitive psychology, which embraces the same kind of quantitative approach illustrated by Hughes’s analysis but sets it in a more developed context. Fred Lerdahl’s Tonal Pitch Space (2001) builds on the foundation of his well-known work with Ray Jackendoff (A Generative Theory of Tonal Music, 1983), but fills in some of the gaps of the earlier theory by incorporating a model of tonal “pitch space” that goes back in its essentials through Schoenberg and Riemann to Öttingen — a twodimensional matrix whose axes are minor thirds and perfect fifths. In essence Lerdahl uses this pitch space as a means of evaluating the perceptual distance between different pitch classes, and this enables him to develop a range of quantitative measures for tonal tension and the attraction between pitch classes and chords, including a model of key derivation. What is important here is not the undoubted technical sophistication of Lerdahl’s model, but the diversity of its linkages. In the first place, like A Generative Theory of Tonal Music, it is based in musical intuition (and Lerdahl is strikingly keen to ground it in his credentials as a composer rather than as a theorist, writing in the very first sentences of the book that, following the publication of A Generative Theory of Tonal Music, “fresh theoretical ideas began to intrude on my time for composing. The only way to unburden myself of them was to work them out and write them down” [2001: v]). In the second place, it not only draws on a variety of established theoretical traditions but is also illustrated through a variety of substantial analytical applications. And in the third place, it is expressed in more or less empirically testable terms, that is to say in terms of predictions of what listeners will perceive—and indeed the development of Lerdahl’s thinking between 1983 and 2001 in part reflects his collaboration with a number of experimental psychologists, in particular Carol Krumhansl. It is these connections which provide the context that was lacking in Hughes’s work. Krumhansl has been a key figure in the interaction between music theory and cognitive psychology to which we referred; other major theorists with whom she has worked include Eugene Narmour, whose theoretical approach to melodic structure she has tested experimentally, concluding that “the uniformity with which the pres-

7

8

EMPIRICAL MUSICOLOGY

ent results supported the implication-realization model encourages the view that the model has successfully codified psychological principles governing melodic expectations” (Krumhansl 1985: 78). In fact she organized a year-long seminar at Stanford University during 1993–1994 in which both Lerdahl and Narmour took part, along with the music theorist Robert Gjerdingen, as well as the music psychologists Jamshed Bharucha, Caroline Palmer, and of course Krumhansl herself; the outcomes were published in a special issue (1996) of Music Perception. All six researchers worked on the first movement of Mozart’s Sonata in E
, K 282, but the result is in some ways reminiscent of the symposium on Schubert’s Op. 94 no. 1 in which Hughes’s article appeared. Of the music psychologists, both Krumhansl and Palmer engage fully with the music theorists’ work, each of them testing predictions derived from Lerdahl’s and Narmour’s models (with generally rather mixed success); Bharucha at least references the work of all three theorists. Lerdahl’s contribution, which anticipates central elements of his 2001 book, includes a final section called “Connections” containing a single sentence on Gjerdingen, and a more extended discussion of Narmour: his aim here is to show how the basic insights of the implication-realization model can be accommodated within his own, more wide-ranging model. Narmour makes singlesentence references to each of Lerdahl and Gjerdingen. As for Gjerdingen, he conspicuously omits any direct citation of either Lerdahl or Narmour (though he does include a pointed reference to “oversimplified assertions based on an imagined calculus of imagined tonal forces,” Gjerdingen 1996: 370), instead contributing an exercise in historical musicology that has few if any points of contact with the other articles. A final contribution by Leonard Meyer (who was not present at the seminars) underlines the effect of fragmentation through being structured as a series of separate responses to each of the participants. The intention of these comments is not to question the significance of such work in its own terms, but to differentiate it from the “empirical musicology” proposed in this book. One obvious point about it is the division of labor: Lerdahl’s, Narmour’s, and Gjerdingen’s work is intrinsically no more empirical than a great deal of music theory, while the work of the psychologists is not, and is not intended to be, musicological (thus Krumhansl specifically writes in the Preface of her Cognitive Foundations of Musical Pitch [1990: vii] that “the approach taken is that of cognitive psychology”). The model is rather one in which music theorists develop their ideas on a more or less intuitive basis, following which they are passed along to the psychologists and tested experimentally (a model replicated in the structure of the special issue of Music Perception, which presents the work of the three theorists followed by that of the three psychologists); in principle the expectation might be that the theorists would then revise their models in light of the experimental findings, though in practice examples of this are rather hard to find. But the more fundamental lack of communication is between the theorists, and the reason for this lies in the nature of the theories. Huron (1999) draws a distinction “between those theories that claim to usurp all others, and those theories that can co-exist with other theories”; in essence Lerdahl’s and Narmour’s theories fall into the former mould (which is why Lerdahl has to translate Narmour’s concepts into his own theoretical language in order to accept them). Another way of saying more or less the same thing is that such totalizing, mutually incommensurable theories place the emphasis less on the

Introduction

analysis as such than on the theory which the analysis serves to illustrate; there is a real sense in which Lerdahl’s and Narmour’s books are not so much about tonal pitch space and melody, but about their respective theories of tonal pitch space and melody. And this theoretical commitment in turn means that the reductions on which their theories are based give the appearance, at least, of embodying beliefs about how the world is rather than simply representing potentially useful strategies for discoveries (to borrow Huron’s words again). Such work, then, has a fundamentally different orientation from the pragmatic, tool-oriented approach to “empirical musicology” presented in this book, whose aim is to document a number of domains in which empirical methods have had an impact on broadly musicological enterprises, to provide some practical guidance in the application of those methods, and to illustrate some examples of the kinds of study that have made use of them, as well as considering some of their theoretical consequences. The book is organized as follows. Chapter 2 (Stock) provides an overview of empirical methods in ethnomusicology with a particular focus on participant-observational fieldwork—a methodology that, though primarily associated with ethnomusicology, nonetheless has considerable potential for musicology more generally (i.e., it has as much application to Gilbert and Sullivan productions, and perhaps classical concerts at the Lincoln Center, as to ritual music in Taiwan). As Stock points out, the central principle of participant-observational fieldwork is the need for the researcher to become, as far as possible, an “insider” in the culture in question, to observe it and participate in it, and interpret it according to its own standards. Since music is very much more than just the production of certain sorts of sounds, but involves a huge variety of processes (social, financial, technological, organizational), participant observation has the potential to confront a researcher with an enormous mass, and wide variety, of data. The aim of the chapter is to provide practical advice on how to conduct and organize this kind of research—from preparation and planning, through the use of a field log, the organization of field notes and audio or video recordings, and appropriate interviewing techniques, to the manner in which the ideas, attitudes, and “expressive styles” of informants are represented in published accounts of the research. There may be considerable barriers to becoming anything like an insider in some specialized musical subcultures (a participant-observer in the Leeds International Piano Competition, for instance, would need to develop formidable skills as a pianist), but as Stock shows, using examples drawn from a wide range of musical traditions, there is a great deal that can be learned by studying any kind of music from within its own cultural practices. Chapter 3 (DeNora) is also concerned with ways in which music might be studied and understood as socially embedded, but from the perspective of contemporary sociological theory and practice. Though the work of Theodor Adorno offers arguably the most ambitious and (still) influential theoretical tradition within the sociology of music, it certainly does not offer much encouragement for adopting an empirical approach to the subject. Adorno’s approach, perhaps best understood against the backdrop of the appropriation of culture for purposes of propaganda in the Third Reich, was firmly rooted in a conceptually sophisticated analysis of music’s ideological dimension; he saw social and ideological structures as replicated within music (as well as acted upon by music), and so encouraged “critical” readings that

9

10

EMPIRICAL MUSICOLOGY

turned away from the actual social circumstances surrounding the production and reception of music, and towards the close reading of musical texts. Like other sociologists of music today (e.g., Martin 2002), DeNora sees serious dangers in the abstraction from social reality that such an approach entails. She identifies the continuing influence of Adorno in the writings of Lawrence Kramer and Susan McClary; such products of the so-called “New” musicology, she argues, maintain the traditional separation between musical works and the contexts of their production, performance, and use, and as a result have no means to describe music as it functions within real social settings, and in specific times and places. By contrast DeNora provides a “toolkit” of different empirical approaches to the analysis of music as a social process, as it participates in people’s everyday lives and sense of identity. She gives examples of work that has examined the impact of social factors on composition (including that of commercial competition on innovation in pop music), social factors in the construction of musicians’ reputations, the relationship between subcultural identity and musical taste, and music’s role in the social construction of subjectivity. These studies use empirical methods ranging from participant observation (as discussed by Stock), through interviews and the analysis of historical documents, to the more impersonal methods of large-scale social statistics and economic surveys. If DeNora’s chapter engages with people’s socially constructed experiences of themselves and others through and around music, chapter 4 (Davidson) considers a range of empirical methods to investigate the social character of music, as seen through the lens of a more explicitly psychological approach. Starting from the observation that the overwhelming majority of music-making is social in one way or another, Davidson looks at empirical methods for investigating music as social behavior, ranging from controlled experiments and personality inventories, through video-recorded observation and covert manipulation of people’s musical environments, to the use of diaries and interviews as ways of tracking people’s involvement in music. The cultural preoccupation with the musical skills of outstanding individuals has led to a significant body of research inquiring into the factors that might explain or predict the appearance of such skills, and Davidson describes the use of large-scale quantitative methods in this field, as well as more focused and intimate enquiries focusing on a single family. This provides an introduction to some of the principles and methods of qualitative data analysis, which are also employed in analyses of the social processes involved in ensemble rehearsal and performance — a domain that has enormous potential but which has only recently been explored within a social psychological context. Finally, a number of authors have proposed that musicians’ social interactions and behaviors are a function of their personality types, and Davidson provides an overview of some of the empirical methods by which people have attempted to measure personality attributes. Performance studies is the area of musicology which has arguably shown the greatest impact of empirical methods, and chapter 5 (Clarke) presents an overview of those methods and influences. As musicology has moved away from its overriding preoccupation with the score, and toward an understanding of music as performance, it has adopted some of the methods—and even some of the explanatory principles—of the kind of research on performance that originated in psychology. Empirical studies of performance go back as far as the end of the nineteenth century,

Introduction

but it is really since the development of cheaper and more powerful desktop computers in the late 1970s and 1980s, with their ability to handle large quantities of data and to record and analyze sound files on hard disk, that detailed empirical performance research has become a practical possibility for a large and growing population of researchers. The chapter gives an outline of this developing pattern of activity, with case studies illustrating the ways in which performance researchers have investigated keyboard performance directly from the instrument, a wider range of performances from sound recordings, and performers’ gestures and body movements from video recordings. Practical advice about handling and interpreting MIDI (Musical Instrument Digital Interface) data is provided, as well as a discussion of the advantages and drawbacks of studying performance from the three most commonly used media (MIDI, sound, and video). Finally, the chapter considers the nature and explanatory function of artificial models of performance, arguing that they are best understood not as attempts to supplant or even mimic human expressive performances, but as a way of establishing the general principles or “norms” against which genuinely expressive performance is projected. Chapters 6 and 7 (Cook and Pople, respectively) move away from music as a cultural, social, or behavioral event, and focus instead on the use of empirical methods in relation to musical scores. Any study of a musical score is empirical in that it pays attention in some fashion to the “data” of the music, but traditional analysis does this only on a very small scale, in relation to single pieces and with very little attention given to the insights that might be gained from a more “data-rich” and comparative approach. Chapter 6 (Cook) is concerned with various ways in which systematic investigations of larger repertories of music can be undertaken, starting with matters of representation (if the aim is to search large databases of music, how do you represent the music in a way that is flexible and appropriate?), and then going on to the kinds of tool that have been developed in order to search for systematic patterns in the data. Following a review of some earlier approaches to score representation and the operations that can be performed on the resulting databases, Cook focuses on David Huron’s Humdrum toolkit—an approach to the representation of scores, and associated search techniques, which aims to be as flexible and open-ended as possible, allowing users to create new tools to suit their own purposes. As examples of the ways in which Humdrum can be used, Cook describes studies in which Huron and his coworkers have evaluated, for example, analytical claims regarding motivic structure in Brahms, the relationship between style and geographical location in folk song, and the extent to which trumpet music is idiomatic (i.e., is designed around the particular qualities and shortcomings of the instrument). Like any software, the practical usefulness of Humdrum depends on availability and usability, and the chapter concludes with an assessment of the difficulties of turning cutting-edge research approaches into the everyday tools of musicological enquiry. Chapter 6 is concerned with the analysis of large bodies of musical data. By contrast, chapter 7 (Pople) takes a closer look at what can be done when systematic methods are applied to individual works—in other words formalized analytical methods. These date back at least as far as the 1960s when, as we have seen, there was a widespread feeling (especially in America) that the analysis of music should

11

12

EMPIRICAL MUSICOLOGY

be made more scientific, but it is really only since the early 1980s — in particular with the publication in 1983 of Lerdahl and Jackendoff’s A Generative Theory of Tonal Music—that serious attempts have been made to formalize, and even to automate, aspects of the analytical process. This formalism is not in itself a development in empirical method, but it has important empirical consequences: once a method is formalized, it becomes possible to apply it in a systematic and uniform manner and from this to discover what the empirical consequences of the approach really are. A range of approaches to which this applies is surveyed (from methods based on artificial intelligence to neural nets), but it is particularly applicable to Pople’s own “Tonalities” project, with which the chapter ends. Designed specifically with a view to the transition between tonality and atonality in the early twentieth century, but with distinctly wider applications, the “Tonalities” software represents an empirical means by which analysts can become increasingly aware of the consequences of their intuitions concerning a piece’s structure. Through testing their intuitions in this way, analysts confront their unconscious preoccupations and blindspots: just as artificial performance models of the kind discussed in chapter 5 confront researchers with the consequences of a given theory, so Pople’s approach uses the computer to flush out the sometimes unwelcome consequences that informal methods can all too easily skirt around. From the relatively clear representational categories of the score, chapter 8 (McAdams, Depalle, and Clarke) turns to the messier empirical reality of musical sounds, and the ways in which those sounds can be represented and empirically investigated. The first half of the chapter deals with a variety of ways in which sound, as a physical signal, can be represented in various kinds of computer-based visualizations, each allowing different kinds of properties to be revealed and different questions to be asked. The chapter provides a detailed and systematic account of the nature of these visualizations and the acoustical principles on which they are based, together with practical advice about how such representations might be generated, and examples from the musicological literature of the use of such methods with a variety of musical styles. The second half of the chapter turns to the ways in which a perceptual (rather than physical) representation of sounds might be used to shed light on musicological questions. Perceptual principles can help to explain how and why sounds group together in both time and vertical texture, which in turn sheds light on a variety of issues involving orchestration, contrapuntal procedure, and rhythmic organization. This discussion of principles, again combined with practical advice about how data might be represented, leads to two final analytical applications—one a perceptually motivated study of orchestration in Schoenberg’s music, and the other a perceptual rationalization of traditional voice-leading rules. Many of the approaches discussed in the book involve the generation of data to which a number of general principles apply, revolving around the conditions under which they are collected and the kind of control that is needed, as well as the identification of optimal forms of data representation and analysis. The purpose of the final chapter (Windsor) is therefore to explain the principles, and some of the specific methods, of experimental design and statistical procedure in a range of musical contexts. Empirical data in broadly musicological research might include any of the following:

Introduction

• • • • • • • • •

Score data Sound recordings MIDI data from keyboards (and possibly other instruments) Video data Diary data from performers, composers or listeners Interview data Audience questionnaires Data from a textual or visual analysis of CD covers or program materials Quantitative data relating to the sale of musical “merchandise” (e.g., recordings, fanzines, tee-shirts)

The value or otherwise of such data lies in the larger research context within which they occur, and the musicological uses to which they are put. Reference to data on tee-shirt sales, for example, may look banal and of only bookkeeping significance—but if research into music and identity, for example, found that a powerful index of audience members’ sense of identity with classical music was their willingness to buy the merchandise associated with it, then it might be important to know that a rise in tee-shirt sales in one specific year of a particular music festival wasn’t just the result of lower prices, fancier designs, or the sudden influx of a new and more style-conscious sector in the audience. Apparently trivial information, in other words, may turn out to be musicologically valuable— but only if appropriately interpreted. The contribution that an empirical approach can make is not to be endorsed (or dismissed) simply because of its empiricism, but rather for what it can help to discover or reveal. And in order to discover or reveal anything at all, we need appropriate methods as well as good questions. That is what this book is about. Notes 1. “The grid through which we permit the figures of resemblance to enter our knowledge happens to coincide at this point (and at almost no other) with that which sixteenth-century learning had laid over things” (Foucault 1970: 22). For a more extended discussion of this quotation and the general issue see Cook 2002: 80. 2. For a characterization of medieval musicology in precisely these terms see LeechWilkinson (2002), chapter 4; for the argument that data-poor fields breed interpretational conservatism see Huron 1999, on which we draw at many points in this chapter. 3. Babbitt 1972b (but written in 1965), 11 – 12; a discussion may be found in Guck 1994. 4. Formally speaking, scientific experiments are designed to refute the “null hypothesis” that no effect is attributable to the factor(s) being tested (you can never prove the null hypothesis).

References Babbitt, M. (1972a). “Twelve-tone rhythmic structure and the electronic medium,” in B. Boretz & E. T. Cone (eds.), Perspectives on Contemporary Music Theory. New York: Norton, 148 – 179.

13

14

EMPIRICAL MUSICOLOGY

Babbitt, M. (1972b). “The structure and function of musical theory,” in B. Boretz and E. T. Cone (eds.), Perspectives on Contemporary Music Theory. New York: Norton, 10–21. Babbitt, M. (1972c). “Past and present concepts of the nature and limits of music,” in B. Boretz and E. T. Cone (eds.), Perspectives on Contemporary Music Theory. New York: Norton, 3– 9. Balzano, G. (1987). “Measuring music,” in A. Gabrielssen (ed.), Action and Perception in Rhythm and Music. Stockholm: Royal Swedish Academy of Music, 177–199. Boretz, B. (1970). Meta-variations: Studies in the Foundation of Musical Thought. Ph.D. dissertation, Princeton University [published serially in Perspectives of New Music 8–11]. Boretz, B. (1977). “Two replies.” Perspectives of New Music 15/2: 239–242. Cook, N. (2002). “Epistemologies of music theory,” in T. Christensen (ed.), The Cambridge History of Western Music Theory. Cambridge: Cambridge University Press, 78–205. Dahlhaus, C. (1983). Foundations of Music History. Cambridge: Cambridge University Press. Forte, A. (1973). The Structure of Atonal Music. New Haven: Yale University Press. Foucault, Michel (1970). The Order of Things. London: Tavistock Institute. Gjerdingen, R. (1996). “Courtly behaviors.” Music Perception 13: 365–382. Guck, M. (1994). “Rehabilitating the incorrigible,” in A. Pople (ed.), Theory, Analysis and Meaning in Music. Cambridge: Cambridge University Press, 57–73. Hofstetter, F. T. (1979). “The nationalist fingerprint in nineteenth-century chamber music.” Computers and the Humanities 13: 105 – 119. Hughes, M. (1977). “[Schubert, Op. 94 No. 1:] A quantitative approach,” in M. Yeston (ed.), Readings in Schenker Analysis and Other Approaches. New Haven: Yale University Press, 144–164. Huron, D. (1999). “Methodology: On finding field-appropriate methodologies at the intersection of the humanities and the social sciences.” 1999 Ernest Bloch lectures, University of California at Berkeley, #3 [accessible at http://dactyl.som.ohio-state .edu /Music220/ Bloch.lectures / 3.Methodology.html]. Kassler, M. (1967). “A Trinity of Essays.” Ph.D. dissertation, Princeton University. Komar, A. (1971). Theory of Suspensions. Princeton: Princeton University Press. Krumhansl, C. (1985). “Music psychology and music theory: Problems and prospects.” Music Theory Spectrum 17 (1985): 53 – 80. Krumhansl, C. (1990). Cognitive Foundations of Musical Pitch. New York: Oxford University Press. Leech-Wilkinson, D. (2000). [Review of] Margaret Bent and Andrew Wathey (eds.), Fauvel Studies: Chronicle, Music, and Image in Paris, Bibliothèque Nationale de France, MS français 146. Journal of the American Musicological Society 53: 152–159. Leech-Wilkinson, D. (2002). The Modern Invention of Medieval Music: Scholarship, Ideology, Performance. Cambridge: Cambridge University Press. Lerdahl, F. (2001). Tonal Pitch Space. New York: Oxford University Press. Lerdahl, F. & R. Jackendoff (1983). A Generative Theory of Tonal Music. Cambridge, Mass.: MIT Press. Lomax, A. (1968). Folk Song Style and Culture. Washington, D.C.: American Association for the Advancement of Science. Martin, P. (2002). “Over the rainbow? On the quest for ‘the social’ in musical analysis.” Journal of the Royal Musical Association 127: 130 – 146.

CHAPTER 2

Documenting the Musical Event: Observation, Participation, Representation Jonathan P. J. Stock

Empirical approaches have contributed to research in the field now named ethnomusicology since at least the nineteenth century. Comparative musicologists and folklorists in Europe, the Americas, and elsewhere each drew on the new technologies of sound recording and mass publication in order to develop distinct forms of scholarly empiricism. More recent generations of folk music scholars and ethnomusicologists have measured their work against further empirical norms, and there remains today an intriguing relationship between the discipline’s research methods, available technology, and the pattern of its truth claims. Ethnomusicologists have developed empirical approaches to transcription, the analysis of musical sound, the distribution of instruments and repertory, learning processes, and performance interaction, and further chapters could be written on each of these areas as well as on several others. Nonetheless, this chapter focuses on the topic of participant-observational fieldwork, because fieldwork not only is of central importance to enquiry in ethnomusicology but also can be a powerful research methodology for other musicologists. This chapter looks at means of gathering empirical fieldwork data and associated issues of interpretation, authority, and representation. Before looking in detail at fieldwork, however, I shall discuss broader aspects of empirical work in ethnomusicology by means of a brief historical summary.

Introduction: The Empirical Urge in Ethnomusicology and Its Antecedents The invention of the phonograph in 1877 was almost a precondition for the discipline of comparative musicology as devised by European scholars in the final decades of the nineteenth century. Although some previous studies had used traveling musicians and sets of musical instruments purchased by colonial collectors, the new technology of sound recording made two crucial contributions to the new discipline: first, it allowed researchers to assemble for comparative analysis extensive collections of musical material from all around the world; second, repeated playback permitted the detailed study (and hence the transcription in modified staff notation) of non-Western musical sounds. Pitches were minutely measured and tonal systems postulated from these calculations. Meanwhile, instruments were categorized, not 15

16

EMPIRICAL MUSICOLOGY

by reference to the contexts of their use within any one cultural system, but strictly in terms of their physical properties and performance technique (see, e.g., Ellis 1885, Stumpf 1886, Abraham and von Hornbostel 1909, von Hornbostel and Sachs 1914). Criticism of the musical styles in question was not undertaken; there was no attempt to explain or rank music in terms of its “greatness.” Instead, the close analysis of recorded examples and musical artifacts was ultimately intended to reveal scientific principles underlying all human music making. Scholars of musical folklore were empirical in a different sense: not for them the careful weighing and measuring of recordings carried out by the comparative musicologists. Rather, many set out to gather and preserve what they saw as the essential music of their own nation. It was not enough to record such music in a haphazard manner, however, whether by means of notation or the phonograph. Instead, researchers conducted their enquiries in a systematic fashion, and they strove to locate those whom they deemed the most authentic representatives of tradition (Sharp 1954: 119). Whole regions were surveyed (Bartók 1931, 1967); folk singers were questioned about their entire repertories; recorded and transcribed materials were deposited in archives—the very real museums of musical works; variant melodies and song texts were tracked down and classified by means of increasingly complex systems (Bronson 1949, 1959, C. Seeger 1966); and national song collections were published and disseminated through educational channels. Nevertheless, the appeals to empiricism articulated by comparative musicologists and folklorists alike were open to challenge. The comparativists’ detailing of cents, 100 to the semitone, and production of “weighted scales” (diagrams that illustrate the relative durations allotted to each musical pitch within a particular extract), look admirably open to replication by any competent scholar armed with the same sound recording and analytical gadgetry. Yet even setting aside the real challenges of deducing a series of fixed, measurable pitches from the continuous sonic fluctuations of live performance (Schneider 1991), it is now clear that the quest for an underlying science of music often overlooked indigenous systems of musical theory and practice, resulting in comparisons between musical apples and oranges. Likewise, the folklorists’ insistence on “authentic” forms, careful tracing of variants (techniques borrowed from manuscript studies in the fields of musicology and literature), and the laying out of melodic and rhythmic types, gives the appearance of contributing substantively to the gradual accumulation of knowledge about the repertory as a whole. However, more recent research recognizes the longstanding changeability of most folk musical traditions. Today’s authentic example has all too often proven to be the previous generation’s radical innovation. Terms such as authenticity, once so readily affirmed in discussions of sampling and representativeness, now have little credibility except in commentary on specific individuals’ or groups’ claims of ownership or identity. If the goalposts of earlier, pre-ethnomusicological empiricism were uprooted by later generations, however, it was only to shift them for replanting elsewhere. First so entitled in the early 1950s, the discipline named “ethno-musicology” (the hyphen was subsequently dropped) represented both a new beginning and a fresh attempt to address existing problems. Comparativist approaches that emphasized the scien-

Documenting the Musical Event

tific study of physical sound were sustained under the new regime by researchers such as Charles Seeger who attempted to develop electronic transcription devices. These tools, among which the sonograph is perhaps the most widely used, were intended to facilitate the empirical study of non-Western musics through sidestepping the human subjectivity that might compromise aural transcription and analysis.1 In a separate yet somewhat similarly inspired development, folklorist Alan Lomax (1968) devised a system of song analysis that tabulated some 37 technical traits of selected vocal examples (phrase length, melisma, tempo, nasality, and so on). These musical features were then compared to cultural aspects of the society in question. This system, cantometrics, allowed the detailed comparative study of songs from many societies without the prior translation of recorded sound into the potentially misleading conventions of Western staff notation (see Figure 2.1). Again, Mantle Hood (1971: 123–196), among others, proposed a new system for the classification of musical instruments: his graphic system of “organograms” was designed to accommodate multiple playing techniques, unlike the traditional hierarchical schemes based on performance technique (the high-level subdivision of lutes into bowed and plucked categories, for instance, made it hard to place an instrument like the violin that could be performed both ways). Ultimately, however, these attempts to finetune existing modes of empirical enquiry have proven tangential to the main thrust of the new discipline since the 1970s. It was the adoption of anthropological fieldwork as a primary research model that marked out ethnomusicology as distinct from earlier comparative approaches. In its classic sense, this meant at least one full year’s residence within the culture being studied, speaking the local language and learning to perform local musical traditions. Extended in situ research was intended to allow sufficient time for the ethnomusicologist to become adept at understanding the contexts within which music making occurs, and to get to know individuals deeply enough for real discussion to take place. Formal interviewing, although sometimes the only way of speaking with particular individuals (and admittedly efficient for checking uncontested details), was seen as the least desirable form of research interaction, and the least likely to generate valuable insights. (Possibly only the mass questionnaire was regarded with greater distrust.) 2 Instead, the ethnomusicologist was to take part in and observe whatever music-related behaviors occurred customarily, becoming part of the existing “community of speech” whose norms and values could be so easily displaced or closed off by the externally imposed constraints of interviewing. Learning to perform was thus seen as a valuable research technique in that it facilitated access to other musical individuals and situations. It also provided personal experience of performance that could generate further research questions and insights. Mantle Hood (1960) coined the term “bi-musicality” to emphasize the researcher’s duty to acquire some level of competence as a practical musician within the cultural setting in question. Nonetheless, and despite the fieldworker’s efforts to develop his or her own musical experience, the new ethnomusicologists agreed that interpretative weight was to be given to the musical explanations and evaluations offered by “insiders,” that is to say members of the society in question: this constitutes “ethnomusicology” in the

17

18

EMPIRICAL MUSICOLOGY

Figure 2.1. Cantometric profile of an excerpt from huju (traditional Shanghai opera) sung by Yang Feifei.

original sense of the term (the musicology of the people), a sense that relates it to ethnopoetics and ethnohistory. Such insider accounts, for which the anthropological term is “folk evaluation,” became the authoritative basis of a new empiricism, and the key responsibility of the fieldworker was to ensure that they were not misappropriated in subsequent written analysis. There is, thus, a sense in which self/other distinctions were both collapsed (the ethnomusicologist personally learns the music and the culture that goes with it) and sustained (the ethnomusicologist cedes evaluative

Documenting the Musical Event

and critical authority to local voices). In sum, then, the new methodology could claim to be empirical in that it was based on sustained observation of and participation within the culture in question, interpreted according to indigenous standards.3

Investigating Music Through Participant-Observation and Fieldwork The study of music through fieldwork may be useful to students of Western musical culture in at least two respects. First, it is self-evident that music is more than simply sets of sounds (and the notating of these sounds in symbols). Music is process as well as product, an arena for both social action and personal reflection; it is “emotion and value as well as structure and form” (A. Seeger 1987: xiv). Clearly, many of these essential parts of the whole complex that is music in any concrete setting are not immanent in the sonic material itself. Rather, they inhere, often only temporarily, among particular groups of socially and historically situated individuals. A study of these aspects of musical life will therefore need to integrate close examination of sound structures and symbols with analysis of the patterns of human action and thought that infuse these structures with meaning in specific social situations. Of course, the researcher able to discover what “music” is among a social group at any specific time is then well placed to assess questions of musical value and change within that society. Equally, the musicologist who analyzes what musicians and others actually do on particular musical occasions, and how these individuals explain what they do, is likely to gain enlightening perspectives on the sounds that emerge—there may be differences between theory and practice, for example, that repay close attention. Second, fieldwork-based approaches offer the socially orientated musicologist access to many kinds of music in Western society that have yet to be extensively written about from a historical perspective, and for which a full range of printed scores, published recordings, and other written documentation may not exist. Church bell ringing, musicals, and amateur popular music offer but three examples. In these cases, personal participation (possibly involving the researcher’s performance as part of the musical ensemble) may be not just the best way of gaining access but the only way for the investigator to proceed. Elsewhere, there may be some written materials that can be rounded out through direct personal research. Professional popular music, for instance, is widely written about in journalism and also in several academic disciplines, yet direct interaction with musicians, fans and others provides new perspectives on a range of issues not yet fully addressed in published writings. These points make it plain that the fieldworker is potentially faced with a huge amount of data: the whole field of musical sound (not only those elements recorded in notation), whole repertories of music-related behavior (from constructing and learning instruments to concert-going practice, modes of shopping for CDs, the use of hi-fi equipment, and habits of in-car listening as well as actual performance), and the social and mental arenas within which equally diverse individuals conceptualize and negotiate aspects of their musical experience. Information within these separate categories may conflict. The members of an ensemble might claim (and believe) that interpretative decisions are made democratically during rehearsal while observation

19

20

EMPIRICAL MUSICOLOGY

reveals that one individual normally provides most of the direction. Just as theoretical principles may be rarely realized in practice but are still valuable as ideals, so too there may be areas of practice that are not readily talked about. And as in the formal interview, the presence of the investigator may itself become a factor that affects the quality and nature of material collected. Gathering, organizing, and interpreting this heterogeneous body of data with any degree of confidence requires careful consideration of a number of problematic issues.

Prefieldwork Preparation The first stage in any piece of field research is preparation. Depending on the site of the research, preparation may include language learning, preliminary instrumental or vocal lessons, library or archival research (familiarizing oneself with the musical traditions in question as well as with existing research), theoretical investigation of potential research questions (including looking at similar projects carried out elsewhere), acquisition of technical skills (learning to use, say, a digital audiotape [DAT] recorder or video camera), writing applications for funding, establishing initial contacts within the field, and assembling required materials (DAT cassettes, film, notebooks and medicine, assuming the research is to be conducted in a location where these may not be readily available). A useful technique at this stage is to draw up a plan of research. The plan should sketch the main questions of the research, list known resources, summarize factors that will affect the conduct of the research, and note other conditions that will have to be met for the research to be satisfactorily conducted (see Table 2.1). Devising a plan makes it possible to cross-relate available resources and conditions in light of the main research questions. Potential problems in the plan might be exposed at once: the plan in Table 2.1 arguably involves so many genres that the researcher is unlikely to be able to acquire real expertise across the whole range. On the other hand, its embracing of instrumental music, vocal music, and dance together with amateur and professional performance contexts may be valuable, and worth sustaining in some form. Perhaps research into two contrasting genres can occur throughout the year with others becoming topics of study for a briefer period at the end, when the researcher has built up more experience. A second criticism might be that although the field looks like being a relatively easy one to gain access to as a participant and observer, its very informality may make it difficult to penetrate deeply. If musicians meet just once weekly and then spend most of that time playing, there may be relatively little opportunity for sustained musical discussion. The researcher may need to try to find another way of meeting some or all of the performers —individual instrumental or vocal lessons, if this is an appropriate model of learning within this community, might be one (though it raises further cost implications). Visiting people in their homes to review recordings or try out instruments might be another; this will require getting to know people well enough to set up such visits and to judge whether one is unreasonably imposing on them. In contexts where initial access is more difficult, the preparatory stage will include a potentially quite extended phase of writing to or telephoning contacts and intermediaries. Before doing so, it is important to discover potential objections and

Documenting the Musical Event Table 2.1. Sample plan of research: vernacular musics in South Yorkshire. Main research questions

Resources

1. What role(s) does folk music play in contemporary England? 2. Who is involved in these musics, and why—what’s in it for them? 3. What is the role of memorization, composition, and improvisation in these musics?

1. South Riding Chamber Orchestra—every Wednesday at the Red House, Sheffield; uses notation 2. Clog dance group 3. Hathersage carol singers—active only pre-Christmas 4. Listings of folk club sessions in SRFN News and local papers

Significant factors

Conditions to be met

1. A well established network of local enthusiasts and scholars; much advice readily available, much material already collected and existing contacts 2. An openness to newcomers both social and technical—no need to be of professional standard in order to take part 3. One year available for main body of field research

1. Easy to meet people collectively at practices and sessions, but will it be possible to talk more deeply in these situations? Will musicians mind being recorded or filmed? (Should I pay for these?) 2. Will it be possible to participate in each of these genres? (Can outsiders join the carol group?) 3. Will it be necessary to memorize large numbers of pieces and songs in order to progress beneath the surface? If so, is there time for this within one year? 4. Transport costs. Other costs? (Beer money!)

think through solutions to them, stressing ways in which the proposed research might advantage those under study (if nothing else, by virtue of an extra pair of hands on site at no cost—some reciprocation for the learning gained is often only fair, even if it is not always demanded). Education staff at a regional professional opera company in the North of England, for instance, were worried that a student who had proposed a period of research into the collective enterprise that constitutes opera production would need constant attention, thereby distracting them from their own work. Nonetheless, they were familiar with work placement schemes, and were supportive of the idea of an initial month-long period of research, a phase during which the student could attempt to gain trust and set up a longer residence. Similarly, school teachers may need to clear an incoming researcher’s presence and activities with a head teacher and parents. Certain groups may simply (and perhaps reasonably) object to the idea of being “studied” at all (in such cases it is considered unethical to investigate them covertly), or it may become clear that, to be effective,

21

22

EMPIRICAL MUSICOLOGY

the research will demand linguistic skills or a commitment of time to on-site residence that the researcher cannot meet. When problems of access or viability cannot be resolved—and courteous persistence often pays off — it may be necessary to abandon the initial research plan and look for another topic altogether. A final aspect of preparation concerns technical familiarization. The fieldworker may not have operated a video camera before, or may be proposing to record performances or discussions on an unfamiliar DAT, mini-disc, or cassette recorder. He or she may be unsure of microphone placement, selection of recording levels, or typical battery life. In such cases, rehearsal and practice are indispensable.

Field Log and Field Notes Once a topic has been selected, its viability examined, and funding secured, the field research itself can begin. At this stage, the fieldworker will need to tailor his or her approach to the musical field and context in question, a context that includes the research subjects’ assumptions about the researcher. The researcher’s gender, age, class, ethnicity, and professional background may encourage the striking of certain attitudes by fieldwork contacts—perhaps they wish to shock, tease, or impress the researcher. While the formal interview or questionnaire is particularly prone to these kinds of distortion, other interactions can easily be affected as well. Two instances will illustrate this observation. A student who started to investigate views on music in gay clubs found that his attempts to initiate informal conversation about the music with individuals outside his group of friends were regularly assumed to be a subtle form of pickup line. Similarly, in educational contexts the incoming researcher may be assumed to be a “teacher,” with the result that pupils try to give the “right” answer, or, more mischievously, impose their own agenda. A second student admitted that he had himself some years earlier been a research subject in an investigation into talented young musicians at a specialist music school. Neither he nor his classmates had taken the research project seriously, he claimed, and they had vied with one another in faking data about the amount of practice time they put in. Fieldwork researchers seek to minimize these kinds of problem through their reliance on participation and observation as well as speaking with “informants” whom they come to know well. (Some writers prefer friendlier terms: consultant or field colleague.) The fieldworker attempts to participate in the life of the community in question for a sufficient length of time that he or she wins the trust and respect of those under study, and to discover through this process of familiarization how to ask questions or encourage conversation leading to genuinely meaningful information. Finally, the fieldworker hopes to become sufficiently aware of the values of the researched community that he or she can produce interpretations or analyses that are well founded. So, in the first example above, it may be that the researcher needs first to expand his circle of contacts among those who attend gay clubs; he needs to gain a clearer idea from experienced regulars about how to approach strangers without being misunderstood, and how best to steer conversation around to musical matters—and this may enable him to reflect more directly on his own musical responses as a listener and dancer, particularly as these develop over time, and to make detailed observations of the responses of others. In the second example, participant-

Documenting the Musical Event

observation might mean becoming a pupil at the music school in question for one academic year (more viable for a younger researcher than an elderly one, perhaps, though the mature student is no more “foreign” than the overseas ethnomusicologist in many another research situation); training (and hiring) one or more pupils as research assistants, thereby involving them personally in the project as researchers and not simply as data-bearers; or taking on a role as teacher in the school in order to gradually win the respect and fuller cooperation of pupils.4 Whatever the specific solution, personal participation can often break down some of the “us and them” difficulties that otherwise threaten the integrity of a project. Participation as some kind of peer (albeit an unusual one) typically leads to a level of personal experience that can directly inform research questions. Moreover, the researcher is well placed through participation to observe the actions of other individuals, and to set a picture of what people actually do alongside what they say they do. Whatever pattern of participant-observation is finally selected, a primary technique of planning each day’s research is by means of a research log. Helen Myers suggested that the fieldworker maintain a log in which a plan for each day’s research is sketched on the left-hand page of each pair; the right-hand page is then used to summarize what actually happened (Myers 1992: 40). This is certainly a means of encouraging good planning and a systematic approach. Commonly, the researcher keeps two further notebooks. The first is the kind that easily fits into a pocket or small bag; this is used for jotting down thoughts or observations as they occur during the day, for sketching diagrams (layout maps of ensembles, venues, rituals), and for noting information (names, addresses, lists of photos taken). If it is appropriate to take notes during discussions — and one can readily imagine many situations when pulling out notebook and pen would inhibit unselfconscious conversation—then this notebook may be used for that also. The second notebook is generally a more substantial one. While some researchers carry it around to refer to points, this book is normally used for the writing up of “field notes” (and diagrams) at the end of each day’s research. (Relatively few fieldworkers use a portable computer, largely because of the amount of other equipment they already need to take with them.) Field notes are, ideally, an unedited record of all that happened and was said on each day of research, rather like a diary in many respects. Compiling field notes takes discipline and patience, in that there may be a great deal to write down from even a relatively straightforward day’s interaction and observation. Furthermore, music events often occur in the evening, meaning that writing up typically occurs late at night; though all ethnomusicologists probably do it sometimes, leaving writing up until the following day seems to place a strain on the frailty of the memory. (Our memories may, nonetheless, be better than we fear, and tape-recorded interviews or lessons can obviously be accurately transcribed days after they occurred.) Each day’s entry should be as thorough as possible, and not simply confined to the material pertinent to present research questions, so that future research can make use of these data as well. Some fieldworkers compile an index for each volume of their field notes, and many leave space at the beginning of the book for a detailed table of contents listing date, location, and other key factors or informants by name. A sample extract from one page of my own field notes is illustrated in Table 2.2.

23

24

EMPIRICAL MUSICOLOGY

Table 2.2.

Sample excerpt from field notes: ritual ensemble music in Taiwan.

Thursday 12 August 1999, Evening

Baifushequ, Jilong, Taiwan

p. 33

There are twelve to fifteen [ Juleshe] members present when I arrive, rehearsing a qu [kio in Taiwanese: instrumental piece] I have not heard before. They break off as I come in and immediately ask me to sing Qi cun lian [the simple qu I was set to learn two days earlier]. I feel somewhat uncomfortable about disrupting the rehearsal, but it seems that they are happy to break off. My first attempt goes none too well: I sing the wrong mnemonic (though right pitch) in line two and mispitch the leap between lines four and five. The musicians don’t criticize me. Wang A’nao (one of the two teachers) simply asks me to repeat. Now everyone has stopped to listen. Wang tells me to read from the notation and to beat [da pai] each time there is a “o” symbol in the score. Again, I perform poorly, hesitating sometimes or singing in time but with incorrect mnemonics. Wang (and several of the others) emphasize that I need to practice more, learn how to keep a steady beat, and memorize better than this. Qiu then fetched a suona [traditional oboe-type instrument] from the wall rack and played through the qu twice for me, while I sang along. A second musician joins in the second time. They comment that I seem to know the tune all right but aren’t yet strong enough on the mnemonics to be able to read other pieces in notation well. Deputy troupe leader Ni fetched a can of beer from the fridge for me, a gesture that caused some teasing from younger ensemble members. “Beer! He has beer. How come we don’t have any? Let’s get some too!” (In fact, they didn’t help themselves to any drinks.) Clearly, I am still a guest. Qiu and Lin started to chew betel nut, and discussion moved on to the qulu [kiolo] repertory—music in which an ensemble including stringed instruments accompanies vocal music. Wang explained that this would involve the bringing in of outside musicians. None of them are practising kiolo right now. He encouraged another senior member to sing [the role] Wang Zhaojun but the man declined— he had no voice (meiyou shengyin). Wang mentioned that the [City?] Cultural Department [Wenjian hui / Wenhua Jianshe Weiyuanhui] organizes a class for this in the afternoons but they’ll be on holiday at present. I ask what I might plausibly be able to learn in the remaining six weeks. There is much declaration of this being too little time—one man states that he himself has been studying for fifteen years and has learned almost nothing; they spend four months revising tunes in each guan [kam, mode]. No specific instruction is forthcoming.

In this extract (approximately one third of the entry for that evening session), the notes are an unsystematic mixture of present- and past-tense commentary, with occasional reminders thrown in. (A few explanatory comments have been added so that the passage makes sense to nonspecialist readers.) There is a small amount of direct quotation, though not very much, largely a result of the mix of languages that characterized the session: some musicians spoke Taiwanese, which I am just beginning to learn, others then translated this into Mandarin, with which I am more familiar. Direct quotation would be more common in cases where the fieldworker is properly fa-

Documenting the Musical Event

miliar with the language, in that we are normally concerned with how individuals express themselves, not simply with what they say. Names of individuals are provided, insofar as I knew them at that stage. There are data here on such issues as rehearsal procedures, modes of learning, and informality of interaction within the group, but this particular passage contains few analytical remarks or theoretical reflections. In fact, the writing up of notes is often a process that stimulates analytical thought. When it does, it is a good idea to note these thoughts down immediately (but labeling them clearly as your own ideas and not views that one of the informants offered). In the literature there are a number of further suggestions concerning the production of field notes.5 Anthony Seeger, for instance, suggests six questions that the field researcher might seek to document (1992: 90): 1. What is going on when people make music? What are the principles that organize the combinations of sounds and their arrangement in time? 2. Why does a particular individual or social group perform or listen to the sounds in the place and time and context that he/she / it does? 3. What is the relation of music to other processes in societies or groups? 4. What effects do musical performances have on the musicians, the audience, and the other groups involved? 5. Where does musical creativity come from? What is the role of the individual in the tradition, and of the tradition in forming the role of the individual? 6. What is the relation of music to other art forms? Answers to these questions as recorded in field notes (and subsequently abstracted in analysis) embody empirical information in four respects. First, the information preserved in carefully assembled notes forms a broad palette of personal experience and socially founded perspectives on which specific musical and social interpretations can be based; through participant-observation the researcher not only begins to understand how it feels and what it means to engage in a particular form of music making, but also is well placed to survey the views of other, probably more expert insiders. Second, this form of in-depth investigation is one in which considerable effort is made to be sensitive to music as made by “real” people in “real” musical contexts. (The quotation marks acknowledge that the specialist denizens and contexts of the university seminar room, philosopher’s den, or psychologist’s lab are real too.) Contextsensitive field research accordingly exhibits scholarly empiricism in its attention to differences between musical processes and musical products, and between musical practices and musical theories—what John Kaemmer (1993: 14) calls practical consciousness and discursive consciousness, terms intended to emphasize that we may be well able to do things we do not habitually talk about (or vice versa). Third, the gathering of field notes is empirical in that it pays particular attention to the separation of the researcher’s views and interpretations from those of the group or society under examination. Value judgments and assumptions on the part of the scholar are normally avoided, with preference given to the (probably multiple and often contradictory) views of informants. We may personally deem The Gondoliers musically superior to Iolanthe, say, but the whole point of field research is to discover what members of the Halesowen Light Opera Group and their supporters

25

26

EMPIRICAL MUSICOLOGY

think, not to impose our views upon them or to use them merely as stooges through whom to put forward our own opinions. When we discover views that are deeply held, we seek to value and respect these, perhaps to analyze or interrogate them, but not to undermine them in our subsequent written studies. Finally, the detailed recording of data in properly maintained field notes — who said or did what, where, when, and under which circumstances — provides a resource that may be returned to regularly in order to develop new hypotheses about music and culture. (Manuals on fieldwork recommend that the researcher reread the entire collection of field notes at regular intervals as a means of stimulating further reflection, picking up on theoretical musings jotted down earlier on, and catching emerging patterns within the data.) Naturally, it is also possible to return to field notes in later years; the society may itself have already changed in the meantime, and subsequent use of the data will need to acknowledge this possibility (or be accompanied by a restudy).

Interviewing Some of the problems of interviewing have already been mentioned. Nonetheless, it is a fact of life that the interview is sometimes the only means of speaking with certain individuals. Given that fieldwork is intended primarily to discover how a certain group of people understand their own music making, it is important that the researcher avoids asking leading questions. In certain communities it is impolite to answer in the negative, which means that yes / no questions are of little use. Elsewhere, employment of long or detailed questions may result in short or unrevealing responses, because the interviewee is not allowed space to formulate answers in terms of his or her own categories of thought—categories that may, in the end, prove more important than the specific facts fitted into them. Furthermore, the interviewer’s perceived role as a “music expert” may lead interviewees to see the experience as akin to a test, with the result that they feel uncomfortable or struggle to second-guess the “right” answer. David Reck, Mark Slobin, and Jeff Todd Titon (1996: 514 – 515) illustrate two sample interviews with the same “consultant” in their advice to new fieldworkers: the first fieldworker’s questions, it can be seen, imply that particular answers are expected, and thus close off reflection on the part of the interviewee, whereas the second fieldworker keeps the discussion going by asking open questions. FIELDWORKER 1: Did you get your first flute when you were a girl? CONSULTANT: Yeah. FIELDWORKER 1: What was the name of your teacher? CONSULTANT: Ah, I studied with Janice Sullivan. FIELDWORKER 1: When was that? CONSULTANT: In college. FIELDWORKER 1: I’ll bet you hated the flute when you first started, I can remember hating my first piano lessons. CONSULTANT: Yeah. FIELDWORKER 2: Can you remember when you got your first flute? CONSULTANT: Yeah. FIELDWORKER 2: Could you tell me about it?

Documenting the Musical Event

CONSULTANT: Sure. My first flute—well, I don’t know if this counts, but I fell in love with the flute when I was in grade school, and I remember going down to a music store and trying one out while my father looked on, but I couldn’t make a sound, you know! FIELDWORKER 2: Sure. CONSULTANT: So I was really disappointed, but then I remember learning to play the recorder in, I think it was third grade, and I really loved that but I didn’t stick with it. Then in college I said to myself, I’m going to take music lessons and I’m going to learn the flute. FIELDWORKER 2: Tell me about that. CONSULTANT: Well, I had this great teacher, Janice Sullivan, and first she taught me how to get a sound out of it.

Film and Sound Recording There is less to say about film and sound recording than field notes, partly because most musicians are already experienced users of cameras and tape recorders, partly because quickly changing technology equally rapidly renders detailed technical advice obsolete (other than the obvious “know your tools”), and finally because the principles that inform the use of film and tape in the field are similar to those that underlie the preparation of field notes. Nonetheless, two aspects of using these media will be briefly discussed here: the special contribution that they may make as research tools in themselves, and issues of preservation and documentation. Turning our attention first to film, it will be immediately apparent that research into music making in its context is likely to be stronger when the researcher pays attention to the visual dimension of the performance event. Multimedia genres such as opera and ballet offer obvious examples, but almost all musical performance makes meaningful use of location, space, and movement, whether in the form of the spatial deployment of organ and choir in a cathedral, the choreographic hand shifts of Chinese qin zither performance (Yung 1984), audience – performer interaction in Pakistani qawwali devotional singing (Qureshi 1995), the processing of competing brass bands through a Northern English town, or the sending outside in Taiwanese beiguan rehearsals of beginners who practice under the watchful tutelage of an experienced musician (see Figure 2.2). Indeed, it is worth stressing again here the unnecessary restrictedness of the pervasive notion within much of the Western academic community that music is specifically a sonic art. If, apparently like most people outside the academy,6 we were to assume in our studies that the visual, motional, and emotional aspects of music are just as central as its aural dimension, we would have taken a step toward constructing a musicology well placed to comment on music as social practice. Video and photographs provide an extra swathe of data, allowing the researcher to look repeatedly and in detail at aspects of the music event that might never be captured in notation. Sometimes these extra data are clearly integral to the cultural practice. Musicians in the Durham-based quintet Juke Box Jive, specialists in the historically informed performance of 1950s–1960s rock, recognized (and demonstrated during performance) the importance of the visual aspect: other than haircuts, clothes, and replica instruments (but replica only visually, as an up-to-date sound

27

28

EMPIRICAL MUSICOLOGY

Figure 2.2. A beginner beiguan player, assisted by Mr. Qiu, practices outside the main Rehearsal Hall (Baifushequ, Jilong, Taiwan; September 3, 1999; photograph by C. Chou).

system was employed), their dance moves, energy of performance, and — as they stressed in discussion—youth were all part of the 1950s image. (In an irony the quintet understood very well, a genuine 1950s rocker would nowadays appear false, since he’d look too old.)7 Moreover, study of these nonaural dimensions of performance can reveal the influence they exert on sound structures: for instance, Regula Qureshi (1995: 148–174, 181–186) was able to map (on diagrams she named “videographs” and “videocharts”) moments where direct audience encouragement led qawwali musicians to offer additional stanzas of certain songs, and where the perceived lack of audience interest led the ensemble to move quickly onto different strains. It is not just that interaction among those present shapes sound structure in certain circumstances, but that musical events can actually create, albeit temporarily, concrete models of social structure through which musically situated individuals move. Examples range from many forms of ritual through to secular dance-centered events: one does not have to be a social psychologist to find social structures in the forming of pairs who move (more-or-less) as one on such occasions. Unlike field notes, video and photographs can also be conveniently viewed by assembled members of the researched community, thereby encouraging further reflection and conversation on the event by those who were involved, and opening up further informed interpretations to the researcher. (Sound recordings can often be profitably used in this way, too.) This is a particularly useful fieldwork device, in that it leads conversation around to what is significant in performance without the imposed formality of the interview. Video and photography also offer a strong means of sharing ideas with subsequent audiences. In that it is normally the researcher who edits the film, selects the photographs, and composes a script to accompany these

Documenting the Musical Event

images, however, issues of representation pertinent to all social research are particularly marked with regard to the use of film and photography, and I shall return to this in the final section. Many researchers place the recordings they make (or copies of these) in archives, so that they can be used by others, including members of the researched community. Table 2.3 summarizes the information normally filed with any archived recording (this is in addition to information recorded in the field notes); while each archive may have its own standard formats for noting this material, it is probable that each will require information within the 10 categories laid out here. Much of this information will have been gathered by the fieldworker and entered into field notes, but a special effort will generally have to be made with regard to the question of copyright and permissions. This can be a particularly thorny problem in contexts outside the jurisdiction of the researcher’s legal system: for example, religious ritual participants in communist China might well prefer no one to write their names down (as might street musicians in certain locations nearer home). A student attended an ethnic drumming course in Britain where supplies of a mild (yet illegal) inhaled narcotic were shared around to encourage collective relaxation; while names can be changed in any published account that happened to mention this detail, the showing of the group on video (and their identification in the accompanying documentation) might break a tacit confidence.

Coda: Representing Other Voices This last example has led directly to the issue of representation in social research. The ethnomusicologist and any other music researcher interested in speaking about “other” people faces a potentially complex web of ethical pressures and agendas. Clearly, we have a responsibility to those whom we represent in our writings (and films, lectures, and other offerings). This responsibility is not discharged simply by changing names and places in the published account (although this can help in certain situations) or by specifically asking for permission for all that we do or write (which again can help). Not only are we building our careers on “their” expertise, but also, and particularly in the case of less well-known groups and societies, the scholar’s published writings may become the only sources available, and thereby shape other outsiders’ expectations about the tradition or people in question. Kay Shelemay’s studies of Ethiopian musical traditions suggested to her that the Falasha people (now more commonly known as the Beta Israel), many of whom were seeking evacuation to Israel, might not—as commonly believed — have descended from a formerly unknown Jewish tribe at all: given the potential political impact of this discovery, she had to consider how and where to publish her findings (Shelemay 1991: 136–152). Our role as mediators between our informants and the academic world (and, through textbooks, school pupils) is also one that demands serious consideration. Kofi Agawu (1995) demonstrates how early studies of African music stressed difference to the degree that significant similarities between principles of African and European music making were overlooked. Effects of the construction of African music as “different” in kind included an emphasis on its exoticism (examples

29

Table 2.3.

Sample checklist for data to be preserved for each item recorded

in the field. 1. Video or cassette number. These are typically allotted sequentially throughout the research trip as a whole, often with the year (and perhaps general location and medium) as a prefix: TAIWAN VID 99/01, for instance, refers to a video recorded in Taiwan in 1999. Apart from the obvious convenience of a numerical sequence, numbering tapes in this way allows the tapes to be marked up beforehand (on the tape itself as well as on the sleeve) such that errors are avoided when switching tapes over under pressure later on. This code number can be entered into the field notes when the tape is used. 2. Exact date. It may avoid confusion among some potential users if the month is written in letters, for example “12 May 1992,” rather than numerically “12/05/92.” 3. Exact location. For interview tapes at people’s homes, recording their address enables the researcher to send a copy of the tape to those interviewed. (See also 10 below.) 4. One-line title for the event, such as “Beethoven Quartet Cycle at Clothworkers’ Hall, Leeds.” 5. Technical data: digital or analogue, stereo or mono, PAL or NTSC, noise reduction, duration, etc.. 6. Participants. This may involve a list of names (perhaps annotated with instruments, roles or other duties) or be simply a note of ensembles or communities involved: “Choristers of Durham Cathedral directed by James Lancelot” or “c.200 mourners, most apparently from Manchester’s Iranian community.” When possible, an indication of the size of the group can be valuable. In some cases, a printed concert program can be filed with the recording. 7. Instruments. Unusual or flexible Western ensembles should be summarized and as full details as possible given for all other instrumental groups, including indigenous names as appropriate: “Chou Chien-Er (voice and wood clapper pie), Diu Zai-Hing (four-stringed lute gibei), Pan Lun-Mui (vertical notched flute xiao), Gang Mo-Gen (three-stringed lute samhen) and Cua Tiam-Mo (two-stringed fiddle lihen).” 8. Items, pieces, or other recording content by track (with real-time durations), including mention of language of speech or singing when other than English (or when use of English is itself significant, for instance in Italian opera performance or German punk rock). Cross-reference to page(s) of field notes where these items are discussed in more detail, or a copy of those pages. 9. Recordist’s name and contact details. Other users may need to contact you. 10. Any ethical considerations. Researchers are sometimes permitted to record materials for their own use but not for public broadcast or publication. If this is the case, or if there is doubt about an aspect of the copyright of the performance recorded, reference should be made to special conditions (“for reference use only, no copies to be removed from the library without written authorisation”), to performers (or fieldworker’s) contact addresses, or to a detailed copyright release. Where appropriate, it is helpful to discuss all this with the performers prior to archiving their materials; the rise of digitization and automated remote access to archived materials has reinforced this need. 11. Acknowledgments to funding bodies or similar.

30

Documenting the Musical Event

that sounded Western were seen as unrepresentative); in seeking to establish African music as a topic worthy of attention, scholars went to great lengths to provide interpretations of African musical activity (polyrhythm, “metronome sense,” etc.) that ultimately reveal more about the assumptions of these scholars than they do about African music. Equally, we have a responsibility to our audiences: we are not simply the mouthpiece of our informants. Our training may give us perspectives on an issue unavailable (and possibly uninteresting) to our informants. We may detect inconsistencies in positions taken by individuals or groups, uncover moments where traditions are constructed by historical revisionism or invention, or wish to compare the views of rival groups; Georgina Born’s (1995) analysis of the institutionalization of the Parisian musical avant-garde in the Institut de Recherche et de Coordination Acoustique / Musique (IRCAM) offers one example of a study in which the author subjects the assumptions of the researched community to (highly revealing) critical examination. One means of writing that is widely employed is that of letting the researched speak for themselves, and thereby distinguishing clearly between internal views and the researcher’s own reading of the situation. But the researcher, as author, still selects which voices get to be heard, how much they are allowed to say, and when they speak — so that the use of quotations does not eliminate the issue of representational ethics. (Indeed, some argue that the potential for scholarly misrepresentation becomes all the stronger.) The onus remains on the researcher to find an honest and sensitive solution to the particular representational challenges exposed during the project. Effective use of informants’ voices has been made by Paul Berliner (1994) in his study of jazz; his authorial presence is often limited to introducing each paragraph or summarizing and abstracting key observations from the many musicians he quotes on each new theme. Sara Cohen’s (1994) analysis of indigenous views on what constitutes the “Liverpool sound” offers a second, if less extreme, example (see also Herbst 1997). There is a need, I would argue, for more research on Western musical traditions that seeks to understand how people outside the academy make sense of music. For such study to be effective, traditional academic assumptions about what constitutes music will need to be set aside, the researcher proceeding instead from a sensitive and careful analysis of what those personally involved value, do, and say in real musical situations. Participant-observation offers one well-established means of engaging with these individuals and communities within their own territory and on their own terms. It also offers the student of music a demanding but absorbingly human means of research. Notes 1. A sonograph (or, in America, melograph) is an electronic device that charts along the vertical axis on graph paper the frequencies, amplitude or, potentially, other features of a stream of sound against duration along the horizontal. The machine is able to detect pitches and details that are inaudible to the human ear, and does not interpret the sound into any preconceived tonal or rhythmic system.

31

32

EMPIRICAL MUSICOLOGY

2. Ethnomusicologists do nonetheless interview informants, and a smaller number use questionnaires and other techniques from quantitative sociological research. 3. This empiricism, like that of previous approaches, is not without its own assumptions, and these too have been interrogated in more recent writing. (There is a huge literature on this in anthropology, from which many ethnomusicologists draw; see, e.g., Clifford and Marcus 1986, Geertz 1988, van Maanen 1988, and Jackson 1989.) Most pointedly, perhaps, questions have been asked about the empirical nature of evidence about “other societies” the music researchers themselves help to create (through discussion, recording, and performance, among other means), and then reformulate into a final written account that employs rhetorical devices and theoretical constructs with definite roots in the Western academic tradition. 4. Henry Kingsbury used the first model of research at two American music schools; see Kingsbury 1988, and compare this with Bruno Nettl’s more generalist account of these institutions (1995). 5. Several ethnomusicological authors have reflected on the processes, rhetoric, and politics of writing in the field (see, e.g., Barz and Cooley 1997). Fieldwork is, of course, a key research mode in most of the social sciences, including folklore, geography, anthropology, and sociology, and writers in each of these disciplines have written about the preparation of field notes; useful texts from these fields include Briggs 1986, Jackson 1987, Sanjek 1990, and Hammersley and Atkinson 1995. 6. Students of mine over the past six years have as a class exercise each spoken with five so-called “nonmusicians” to attempt to discover these individuals’ definitions of music. With a total sample of nearly 1,000 people who have answered these questions, it is perhaps significant that a higher proportion have used words like “identity,” “feeling,” and “mood” than have referred to “sound” or related sonic terms. 7. In proper ethnomusicological-cum-empirical fashion, this footnote gives the date and location of the specific meetings at which these observations were made: Juke Box Jive in performance (November 14, 1997, Sunderland) and personal communication, November 21, 1997, Durham. 8. A revised or invented tradition is, nonetheless, every bit as valuable, all else being equal, as one with an unchallenged historical record: comments about authenticity above can be reiterated here. The point is that a group may not wish their claims to authenticity or ownership to be critiqued at all.

References Abraham, O., and E. M. von Hornbostel (1909). “Vorschläge für die Transkription exotischer Melodien.” Sammelbände der Internationalen Musikgesellschaft 11: 1–25. Agawu, K. (1995).“The invention of African rhythm.” Journal of the American Musicological Society 48: 380– 395. Bartók, B. (1931). Hungarian Folk Music. London: Oxford University Press. Bartók, B. (1967). Romanian Folk Music, 3 vols., ed. B. Suchoff. The Hague: Martinus Nijhoff. Barz, G. F., and Cooley, T. J. (eds., 1997). Shadows in the Field: New Perspectives for Fieldwork in Ethnomusicology. New York: Oxford University Press. Berliner, P. F. (1994). Thinking in Jazz: The Infinite Art of Improvisation. Chicago: University of Chicago Press. Born, G. (1995). Rationalizing Culture: IRCAM, Boulez, and the Institutionalization of the Musical Avant-Garde. Berkeley: University of California Press.

Documenting the Musical Event Briggs, C. L. (1986). Learning How to Ask: A Sociolinguistic Appraisal of the Role of the Interview in Social Science Research. Cambridge: Cambridge University Press. Bronson, B. H. (1949). “Mechanical help in the study of folk song.” Journal of American Folklore 62: 81 – 90. Bronson, B. H. (1959). “Toward the comparative analysis of British-American folk tunes.” Journal of American Folklore 72: 165 – 191. Clifford, J., and Marcus, G. E. (eds., 1986). Writing Culture: The Poetics and Politics of Ethnography. Berkeley: University of California Press. Cohen, S. (1994). “Identity, place and the ‘Liverpool sound,” in M. Stokes (ed.), Ethnicity, Identity and Music: The Musical Construction of Place. Oxford: Berg, 117–134. Ellis, A. J. (1885). “On the musical scales of various nations.” Journal of the Society of Arts 33 / 1,688 (March 27, 1885): 485 – 527. [Repr. in K. K. Shelemay (ed.), Garland Library of Readings in Ethnomusicology 7, 1 – 43. New York, 1990.] See also Journal of the Society of Arts 33 / 1,690 (April 10, 1885): 570 (correspondence from Ellis concerning the article); and “Appendix to Mr. Alexander J. Ellis’s paper on ‘The Musical Scales of Various Nations,’ read 25th March, 1885.” Journal of the Society of Arts 33/ 1,719 (October 30, 1885): 1102 – 1111. Geertz, C. (1988). Works and Lives: The Anthropologist as Author. Cambridge, UK: Polity Press. Hammersley, M., and Atkinson, P. (1995). Ethnography: Principles in Practice, 2nd ed. London: Routledge. Herbst, E. (1997). Voices in Bali: Energies and Perceptions in Vocal Music and Dance Theatre. Hanover, N.H.: Wesleyan University Press and University Press of New England. Hood, M. (1960). “The challenge of ‘bi-musicality’.” Ethnomusicology 4: 55–59. Hood, M. (1971). The Ethnomusicologist. New York: McGraw-Hill. von Hornbostel, E. M., and Sachs, C. (1914). “Systematik der Musikinstrumente.’ ” Zeitschrift für Ethnologie 46: 55 – 90. [Eng. trans. by A. Baines and K. P. Wachsmann in Galpin Society Journal 14 (1961): 3 – 29.]. Jackson, B. (1987). Fieldwork. Urbana: University of Illinois Press. Jackson, M. (1989). Paths Toward a Clearing: Radical Empiricism and Ethnographic Inquiry. Bloomington: Indiana University Press. Kaemmer, J. E. (1993). Music in Human Life: Anthropological Perspectives on Music. Austin: University of Texas Press. Kingsbury, H. (1988). Music, Talent, and Performance: A Conservatory Cultural System. Philadelphia: Temple University Press. Lomax, A. (1968). Folk Song Style and Culture. Washington, D.C.: American Association for the Advancement of Science. van Maanen, J. (1988). Tales of the Field: On Writing Ethnography. Chicago: University of Chicago Press. Myers, H. (1992). “Fieldwork,” in Helen Myers (ed.), Ethnomusicology: An Introduction. London: Macmillan, 21 – 49. Nettl, B. (1995). Heartland Excursions: Ethnomusicological Reflections on Schools of Music. Urbana: University of Illinois Press. Qureshi, R. B. (1995). Sufi Music of India and Pakistan: Sound, Context and Meaning in Qawwali. Chicago: University of Chicago Press. [Originally published by Cambridge University Press, 1986.]. Reck, D., Slobin, M., and Titon, J. T. (1996). “Discovering and documenting a world of music,” in J. T. Titon (gen. ed.), Worlds of Music: An Introduction to the Music of the World’s People, 3rd ed. New York: Schirmer, 495 – 519.

33

34

EMPIRICAL MUSICOLOGY

Sanjek, R. (ed., 1990). Fieldnotes: The Makings of Anthropology. Ithaca, N.Y.: Cornell University Press. Schneider, A. (1991). “Psychological theory and comparative musicology,” in B. Nettl and P. V. Bohlman (eds.), Comparative Musicology and Anthropology of Music: Essays on the History of Ethnomusicology. Chicago: University of Chicago Press, 293–317. Seeger, A. (1987). Why Suyá Sing: A Musical Anthropology of an Amazonian People. Cambridge: Cambridge University Press. Seeger, A. (1992). “Ethnography of music,” in H. Myers (ed.), Ethnomusicology: An Introduction. London: Macmillan, 88 – 109. Seeger, C. (1966). “Versions and variants of the tunes of ‘Barbara Allen’ in the Archive of American Folksong in the Library of Congress.” Selected Reports in Ethnomusicology 1 /1: 120– 167. Sharp, C. J. (1954). English Folk Song: Some Conclusions. London: Methuen. [First published 1907.] Shelemay, K. K. (1991). A Song of Longing: An Ethiopian Journey. Urbana: University of Illinois Press. Stumpf, C. (1886). “Lieder der Bellakula-Indianer.” Vierteljahrschrift für Musikwissenschaft 2: 405– 426. Yung, B. (1984). “Choreographic and kinesthetic elements in performance on the Chinese seven-string zither.” Ethnomusicology 28: 505 – 517.

CHAPTER 3

Musical Practice and Social Structure: A Toolkit Tia DeNora

The sociology of music has a strong empirical tradition, yet retains inspiration from its more philosophically oriented past. For sociologists, especially in recent years as the field has experienced a cultural and interpretative turn, the study of music has been linked to wider questions concerning social structure, stability and change, the interaction between social networks and musical production, the emotions, the body, the study of social movements, identity politics, and organizational ecology. In all these areas, sociologists of music have sought to ground their enquiries through the use of empirical methods designed for the scrutiny of behavioral trends, organizations, and forms of action. In this chapter I take stock of the sociology of music’s “toolkit” and present some of the best-known empirical work within the field. My discussion is organized around two broad areas of study: musical production and musical consumption. To contextualize these topics, and to differentiate the empirical sociology of music from musicology’s growing interest in social constructionism, I begin with a brief sketch of classic, and more overtly theoretical, work in music sociology.

Sociology of Music: The Classic Legacy The most sociologically ambitious theoretical perspective to be developed during the last century is to be found in the work of T. W. Adorno (1903–1969). Adorno’s perspective is distinguished by its comprehensive vision, and for the central place it accords to music within modern (and, as Adorno perceived, often repressive) culture and social formation. In contrast to Max Weber’s more formal concern with the origins of musicaltechnical practices specific to the West (Weber 1958), Adorno focused on the question of music’s ideological dimension. In line with classical philosophers such as Plato and Aristotle, he pursued the question of music’s ability not only to reflect but also to instigate or reinforce forms of consciousness and social structures. For Adorno, different forms of music were homologous with (structurally parallel to, and thus able to inculcate) cognitive habits, modes of consciousness, and historical developments. As he saw it, music’s compositional processes—its degree of conventionality, the interrelation of musical parts or voices, the arrangement of consonance and dissonance—could serve as means of socialization. This ultimately structuralist notion 35

36

EMPIRICAL MUSICOLOGY

is perhaps best exemplified by considering Adorno’s views on the contradictory possibilities for consciousness posed by twentieth-century musical forms. On the one hand, he believed that Schoenberg’s music could enable critical consciousness because, through its processes of composition—for example, its use of dissonance and formal fragmentation—it modeled a mode of critical attention to the world that refused to offer “false” musical comfort. On the other hand, jazz, Tin Pan Alley, and other popular genres inculcated psychological regression and infantile dependency (Adorno 1990; Witkin 1998), providing, in the age of “Total Administration,” a medium that “trains the unconscious for conditioned reflexes”(Adorno 1976:53). “Wrong” music thus had to be denounced, and for this reason, Adorno considered socio-musical study to possess a special urgency: given music’s capacity to “aid enlightenment” (Adorno 1973:15), socio-musical analysis was nothing less than a tool for liberation. These are certainly profound questions and ones to which musicologists are increasingly drawn. During the 1970s, interest in Adorno’s work was located peripherally within musicology (e.g., Subotnik 1976, 1978, 1983; see also Subotnik 1990 and McClary 1991:175n). During the 1980s and 1990s, by contrast, musicologists increasingly turned to Adorno. While they generally rejected his dismissal of popular culture and his notion that truly great, liberatory music was that which had “escaped from its social tutelage and is aesthetically fully autonomous” (Adorno 1976: 209), they took up Adorno’s concern for music’s social and ethical character. In particular, they sought to “ground” musical works, and the values embodied in them, either through showing how musical representations inscribed social relations (Subotnik 1983; Leppert and McClary 1987), or through relating them to a cultural history or psychology of music consumption (Cook 1990, Frith 1996, Johnson 1995). Consideration of how musicologists responded to Adorno, and more broadly of the way in which they adopted a social-critical perspective, helps to illuminate some of the differences between musicology’s and sociology’s “toolkits.” It also provides a springboard into contemporary sociology’s more “action-oriented” focus on music as social practice, its shift away from a homology-centered, structuralist paradigm, and its quite different take on the “social construction” of music. Ten years on from Goehr (1992) and Randel (1992), a form of social constructionism thrives in musicology, one that opposes itself to traditional understandings of what is “natural” in music. Even basic, previously taken-for-granted concepts such as the musical “work” have been deconstructed, shown to be purely social constructions of restricted historical and geographical application. Today, most musicologists would probably agree with Randel’s apt observation that musicology’s traditional “toolbox” was designed for the construction and maintenance of a canon of acceptable topics, namely, works and composers. But, as I shall suggest in this chapter, the forms of constructionism now prevalent within musicology are, from a current sociological perspective, not so different from the structuralism characteristic of Adorno’s work. Although there are some notable exceptions, particularly studies of musical listening, reception and use, constructionist approaches in musicology still center on works, and on critical readings of them that aim to reveal the music’s social content. In the writings of Lawrence Kramer and Susan McClary, for example, we are di-

Musical Practice and Social Structure

rected to see music as structurally similar (homologically linked) to social phenomena, or as a “representation” of some extramusical phenomenon. The methodological toolkit here—uncovering intertextual allusion, identifying conventional tropes and the ideological connotations and functions of these tropes, comparing (some aspect of) music’s structure with (some aspect of) the structure of something else — maintains a separation between works and the actual contexts of their production and reception. While social contexts and contents are the ultimate quarry of this type of “New” musicology (as the work of such writers as Kramer and McClary was termed in the 1990s), they are typically pursued through the analysis of texts, rather than through more ecological, empirically oriented investigations of the production, distribution, and consumption of music. Such a move also sidesteps the contested meanings that arise within particular contexts, for example, through resistance to particular musicological interpretations. In short, it is impossible to specify music’s mechanism of operation: there is no methodology for describing music as it acts within actual social settings, specific spaces, and in real time. I do not here wish to imply that sociology cannot benefit from or be compatible with this type of text-based musicological constructionism; on the contrary, a weakness of sociology has been its failure to deal with music’s specifically musical materials, and here textual interpretation and analysis can help to draw sociological studies on to more firmly musical terrain. Nor do I wish to imply any clear division of labor between musicology and sociology; some of the best “sociological” work on musical topics is currently being done by musicologically trained scholars (e.g., Pasler forthcoming). Rather, I wish to contrast the textual focus of “New” musicology with the emphasis of the sociology of music, particularly since the late 1970s, on an action-based paradigm—one that is concerned with the matrices and milieus in which action is framed and effected. Howard Becker (1989: 282) put his finger on the difference when he wrote, with disarming clarity, that sociologists of his persuasion (generally termed “social interactionists”) “aren’t much interested in ‘decoding’ artworks [but rather] prefer to see these works as the result of what a lot of people have done jointly.” This version of constructionism treats music as a social process, focusing on how musical structures, interpretations, and evaluations are created, revised, and undercut with reference to the social relations and contexts of this activity. It is also concerned with how music provides constraining and enabling resources for social agents—for the people who perform, listen, compose, or otherwise engage with it. As the sociologist Pete Martin (1995: 42) has observed, “in general this ‘turn to the social’ in musicological studies has not led to a sustained engagement with the themes and traditions represented within the established discourse of sociology” — themes and traditions that are at some remove from Adorno and his structuralist perspective. Martin calls instead for a focus on music as it is lived and experienced, quoting the Swedish musicologist/ethnomusicologist, Olle Edström, on how the members of his group at Gothenburg responded to Adorno: “we gradually gained a deeper insight into the pointlessness of instituting theoretical discourses on music without a solid ethnomusicological knowledge of the everyday usage, function and meaning of music” (Edström 1997: 19, quoted in Martin 2000: 42). Edström and Martin both allude here to a shift in focus from abstract theory and “macro” issues (such as systems, societal structures, and norms) to grounded theory

37

38

EMPIRICAL MUSICOLOGY

and “micro” concerns (such as a focus on individual and collective practice). Part of this shift centers upon the concept of social agency, on how both social and musical forms (including meanings) are put together or accomplished jointly, in Becker’s sense. This focus on activity is, as I shall argue intermittently throughout this chapter, a very useful perspective. It is dedicated to elucidating the links between social and musical structures in ways that are more than hypothetical. It conceives of the music-society nexus in terms of the pragmatic contexts within which musical works take shape and come to have “effects” in real situations. This focus on action provides an alternative to homological models and their text-centered methodological toolkit—to the emphasis, pace Adorno, Attali (1985), and Shepherd (1991), on how music “reflects,” “anticipates,” or is structurally analogous to social developments or social structures. From a social-interactionist perspective, then, neither Adornoinspired sociology of music nor musicology’s version of constructionism is sufficient for illuminating (“grounding”) music’s sociality. The problem with both these inherently structuralist, text-centered modes of study is simply this: they are oriented to the recognition of patterns and structural affinities between two or more realms (music and some aspect of society—ideology, gender or class relations, identities, cognitive styles), but they are not able to document the mechanisms that create these patterns, that is, to describe how music informs or enters social life, and vice versa. They assert links between music and society, but their methodological toolkits do not equip them to show these links in terms of how they are established and how they function within actual musical and social contexts. By contrast, newer sociological perspectives concerned with social agency investigate the social processes through which these links are forged. As the French sociologist Antoine Hennion says, “it must be strictly forbidden to create links when this is not done by an identifiable intermediary” (1995:248). By this, Hennion means that while music may be, or may seem to be, interlinked to “social” matters, for example. patterns of cognition, styles of action, ideologies, institutional arrangements, such links should not be assumed. Rather, they need to be specified (observed and described) at their levels of operation, for instance in terms of how they are established and come to act. We need, in short, to follow actors in and across situations as they draw music into (and draw on music as) social practice. And this is where empirical methods come into their own within the sociology of music. There are good parallels and precedents to be found in the social study of another “technical” realm: science and technology, in particular in the study of science-in-themaking (Knorr-Cetina 1981; Latour and Woolgar 1986; Bijker, Hughes, and Pinch 1987; Latour 1987). It should be underlined here that these studies of scientific practice and knowledge formation, most of which have been conducted by sociologists with advanced training in the sciences, have concentrated on action — on the situated production of scientific matters of fact, step by (sometimes contested) step. In this respect, such action-based studies move well beyond more general concerns with the parallels between science and society. And some recent studies of this sort have begun to focus explicitly on music technology and musical culture (e.g., Pinch and Trocco 2002). It is, then, in the focus on culture-producing worlds that the sociology of music has found its empirical feet, and thus a way to ground its claims about the links be-

Musical Practice and Social Structure

tween music and society. More specifically, as I will describe below, such work centers on action: on musical practices in and across musical and extramusical realms. For example, it is concerned with musically engaged actors as they constitute (and negotiate the constitution of) music through performance, through coordination, and through reception. It is also concerned with how these constitutive processes in turn draw upon music to constitute other social realities, realities that may exceed the musical but that may, simultaneously, be articulated with reference to music. And with this focus it is possible to dispense with the music-society dichotomy, and to think instead of musical practice as, inevitably, social practice.

Sociology of Music: Musical Production and Its Milieux During the 1970s and 80s, and particularly in the U.S. and U.K., new paradigms were developed that sought to explore music’s links to social processes and contexts rather than structures. Here, music was conceptualized, simply, as social activity. Known as the “production of culture” approach, and developed by scholars such as Peterson (1976), Wolff (1981), Becker (1982) and Zolberg (1990), this perspective provided an effective antidote to the overly theoretical character of Adorno-influenced models. It reinvigorated the sociology of music in its emphasis on action and action’s matrices. It reconceptualized the composer, or music producer, as a member of a musical world or community, and as working with and abiding by (or reacting against) conventions and work practices in order to make music. This view was deliberately prosaic; the production of culture approach sought to demystify the romantic notion of “the composer” and its attendant ideology of the genius in the garret. Karen Cerulo’s (1984) study of change in musical composition across six countries during the Second World War serves to illustrate these points. Cerulo focused on the social disturbance brought about by war and its relation to music-compositional practice. She examined the prewar and wartime activities of composers whom she divided into two groups, those located in combat zones and those who operated in more stable environments. She began with the hypothesis that the work of composers located in areas most characterized by social upheaval due to war would exhibit most evidence of stylistic change, with composers based in non-combat zones showing less evidence of change in their compositional styles and practice. She established a sample of wartime works, focusing on pieces that were intended by their composers explicitly as reactions to the war, and compared these with prewar works by the same composers so as to identify any changes in style during the war years. Government-sponsored works were excluded, on the grounds that they may have needed to portray official sentiments (through uplifting march rhythms and so on). Thus delimited, Cerulo’s sample consisted of 16 works by 14 composers over six countries—combat zone (wartime England, France, Hungary, Germany, and Russia) and noncombat zone (prewar England and the U.S.). These works were examined in terms of the following features, conceived of as dependent variables (see this volume, p. 219, for a definition of dependent variables): melodic structure, tonality, dynamics, rhythm, medium of expression and form. (“For purposes of pedagogical vividness and ease of exposition,” however, Cerulo’s discussion of her find-

39

40

EMPIRICAL MUSICOLOGY

ings focused primarily on melody.) In particular, Cerulo sought to measure the degree to which melodies were conjunct (“smooth gradations”) or disjunct (“leaping motion”) before and after the onset of war in each zone. She plotted melodic pitches using crotchets —one for each new pitch—so as to achieve a graph of melodic spacing for each work. She concluded that while before the onset of war the works of all composers in the sample—combat zone and non– combat zone composers, exhibited jagged melodies, after the beginning of the war those in combat zones became conjunct and lengthy, while those in noncombat zones remained unchanged (1984: 892). From this, Cerulo concluded that she had found evidence for the impact of disruption on compositional practice. She then turned to the critical question: how was one to explain this apparent shift in compositional practice? While older sociological paradigms might have pointed to a homology (or reverse homology) between disruption in society and conjunction in music, with perhaps an associated psychological explanation of trauma and its impact on composers’ needs for consonance and congruence of musical material, Cerulo took a different and more pragmatic tack. She emphasized instead how war-zone composers were cut off from normal music-world interactions, from information and communication with fellow composers, and from access to music publications: “The loss of contact with peers experienced by Combat Zone composers destroyed their professional community.” This, in turn, Cerulo suggested, “led to the unraveling of the normative prescriptions that govern techniques of composition. Consequently, in the absence of both a supportive system and its enforcement by contemporaries of normative adherence, composers deviate from their current paradigm of musical construction” (184: 900). To be sure, these conclusions may provide a source for fruitful debate by music historians: why, for example, if changes in stylistic practice were a function of loss of normal networks and communication patterns, should the deviation of isolated warzone composers all exhibit the same basic tendency — the shift from disjunct to conjunct melodic lines? How might the study benefit from more detailed consideration of the individual work-lives of composers? Does the graphical method of plotting melodic movement provide a valid means for comparing different melodic structures? Could identification and measurement of the parameters of compositional material be combined with an ethnographic understanding of the meanings (local, regional, biographical) associated with musical materials and practices? I suggest that the value of Cerulo’s work (and the justification for reading it today) lies in her general interrogative strategy, her bold attempt to specify measurement techniques for the study of compositional practice and, in particular, her focus on production networks and communication as a determinant of this practice. Cerulo’s study is important in the present context not only because it was one of the first sociological works to deal with musical forms and stylistic change, but also because it can be regarded as a pivot between the older homological model and the newer approach, with its emphasis on music-producing worlds and on the social contexts of artistic production. As Cerulo (1984: 885) put it: “the limited body of literature dealing with the transection of artistic creation and social structure consists almost entirely of large-scale, speculative theories which are heavily influenced by sociohistorical arguments, and whose illustrative support often rests on the sty-

Musical Practice and Social Structure

listic and structural changes in the music of a single composer, or a particular musical tradition.” While seeking to distance herself from “speculative theory,” Cerulo also set her sights on matters that connected back to the grand tradition within music sociology—concerns that were addressed by the earlier homological perspectives she sought to transcend. On the one hand, her work can be read as in contrast to structuralist approaches, such as Lomax’s (1968) “cantometric” investigation of correspondences between song styles and societal structures. (For Lomax, song styles reflect societal forms and, thus, thus habits of mind congruent with these forms — see this volume, p. 17, for further details.) On the other hand, Cerulo wished to retain Lomax’s concern with musical style and its variation across social space — too often, she argued, ignored by the new perspectives and their focus on production, markets and patronage—while linking that concern with a focus on the production circumstances of composers. In this sense Cerulo’s study represented a pioneering attempt to illuminate the “transection,” as she put it, of structure and creation: that is, to devise means of measuring the impact of a changed social context on creative activity in music. By 1989, the “production” perspective was firmly established in not only the anglophone but also the francophone world, after Pierre Bourdieu’s (1984) work on taste publics and social classification systems, and Bruno Latour’s studies of science worlds and science in the making (1987). These perspectives and the various publications that issued from them drew upon detailed empirical study — ethnographies, cultural and social histories, quantitative surveys, and studies of institutions. It was precisely what Becker referred to as “what a lot of people have done jointly” that formed the focus of sociological investigation between, roughly, 1978 and the middle 1990s. In retrospect, the contributions of these years may be set in one of three broad categories: (1) conditions of production (2) the construction of musical value and reputation, and (3) musical tastes, consumption, and social identity.

Conditions of Musical Production Cerulo’s work is representative of a large number of studies aiming to show how the content of musical works is shaped in relation to musicians’ working conditions. Elias’s (1993) pioneering consideration of Mozart, for example, suggests that Mozart’s compositional scope was hampered by his location between two patronage modes and his inability to escape the shackles of aristocratic control. Similarly, Becker’s (1963) study of dance musicians documents how career patterns and occupational opportunities are shaped by patrons and by the need to find a fit between musicians’ aspirations and tastes and what their publics will tolerate. Not only are individual compositional practices affected by productional organization, but so too is the selection of compositions that are ultimately produced and marketed. Peterson and Berger (1990 [1975]) illustrated this point in a highly influential study that revealed how musical innovation was enabled and constrained by infrastructural features of the pop music industry; their work suggested that innovation in pop arises from competition between large record companies and their smaller rivals, showing that diversity in musical forms (as they are produced and reach their publics) is inversely

41

42

EMPIRICAL MUSICOLOGY

related to market concentration. At the time their article was published, Peterson and Berger were trailblazers for the “production of culture” perspective, and their study still serves as a model of how to conduct work in this tradition. Peterson and Berger examined number one hit songs over 26 years of record production, from 1948 to 1973, dividing this period into five eras of greater and lesser degrees of market concentration. Eras of high market concentration were those in which a high proportion of the annual production of hits was produced by one of the four leading companies: during such eras, these companies controlled over 75 percent of the total record market (in fact just eight companies produced nearly all the hit singles). Peterson and Berger considered whether such concentration bred homogeneity of product, pursuing this question by examining the sheer number of records and performers who recorded the hits during their five eras; the thinking was that there might be little incentive to introduce “new” products under conditions of market concentration. They also examined the lyrical content of hits, tracing these variables through the five eras as competition between record companies grew and then diminished over the 26-year period. Simultaneously, they considered indicators of what they termed “unsated demand,” such as changes in record sales and the proliferation of music disseminated through live performance and backed up by independent record producers—genres such as jazz, rhythm and blues, country and western, gospel, trade union songs, and the urban folk revival. They then considered the conditions under which the independent producers were able to establish more secure market positions as the top four producers lost control of merchandising their products over the radio. Finally they traced how the record industry and its degree of market concentration expanded and contracted cyclically over time. By studying conditions of record production and marketing, relating these conditions to new developments in the communications industry, and examining trends in record output and product diversity, Peterson and Berger concluded that changes in concentration lead rather than follow changes in diversity, and that this finding “contradicts the conventional idea that in a market consumers necessarily get what they want” (p. 156). Their study not only highlighted the impact of productionorganization on musical trends and styles; it also outlined how popular music production is characterized by cycles, and detailed some of the mechanisms that affect cyclic development. Peterson and Berger’s study set the scene from the 1970s onward for the concern, in popular music studies, with the production system. Negus (1992), for example, has suggested that working practices within the popular music industry are linked to an artistic ideology associated with college-educated white males who came of age in the “rock generation” of the 1960s and 70s. This occupational stratification has consequences for the types of pop that are produced: women and unfamiliar styles and artists, for example, are marginalized (Steward and Garratt 1984). Such forms of gender segregation may also be seen in pedagogical settings (Green 1997), particularly with regard to instrument choice —a topic that overlaps with work by social psychologists (O’Neill 1997). In the “production” studies discussed so far, the primary methodological strategy consists of a focus on organizational contexts of musical production, and an at-

Musical Practice and Social Structure

tempt to conceptualize musical work as not so different from other types of work, insofar as it requires collaboration, resources in the form of materials, conventions, and communication. Through this strategy, music’s link to social structure is specified: musical structures are examined in terms of their links to the local contexts or musical worlds in which they are produced, distributed, and consumed. The production perspective thus illuminated the impact of social structure on music in highly concrete ways; it highlighted the mundane circumstances under which musical work gets done, the circumstances under which careers are forged and styles developed and changed. On the heels of the production focus and its attention to creative milieux, came sociological studies of the construction of both musical value and reputation.

The Construction of Musical Value and Reputation The stratification of composers, styles, and genres is a rich seam of socio-musical research. Historical studies have helped to unveil the strategies by which the musical canon and its hierarchy of “Master [sic] Works” was constructed and institutionalized during the nineteenth century in Europe (Weber 1978; 1991; Citron 1993) and America (DiMaggio 1982). Both an aesthetic movement and an ideology for the furtherance of music as a profession, the fascination with “high” music culture during the nineteenth century was simultaneously a vehicle for the construction of class and status group distinction. It was also a device of music marketing and occupational advancement. More recent work in this area has gone beyond the distinction between “high” and “low” musical forms. It now includes the issue of how “authenticity” is constructed and contested (Peterson 1997), dismissing the idea of the “work itself” in favor of particular configurations of the work in and through particular performances (Hennion 1997; see also chapter 5, this volume). And it examines the practices and strategies through which particular versions of aesthetic hierarchies are stabilized. For example, Hennion (1989) has drawn comparisons between the recording studio and the scientific laboratory, showing how musical value and scientific fact are both produced through producers’ liaisons with various groups such as the public and the media. Similarly, Maisonneuve (2001) has focused on the way in which the twentieth-century technology of the gramophone afforded music’s users new and more intensely personal modes of experiencing the love for music. Drawing upon record reviews, catalogues, liner notes, and other documents, Maisonneuve suggests that this technology facilitated a music user actively engaged in constructing her or his tastes and monitoring self-responses. By comparing the two major technological revolutions in music distribution during the century, she shows how both musical listening and the listening subject were technologically transfigured. Her study thus builds upon and gives a new type of spin to William Weber’s pioneering work on the emergence of modern musical consumption and notions of “music appreciation.” Similarly, it highlights the extent to which the consumption of music involves more than listeners and works, consisting also of networks or, as Maisonneuve puts it, “set-ups” of objects, postures, habits, and evaluative discourses.

43

44

EMPIRICAL MUSICOLOGY

Sociological studies of musical value can be regarded as critical or even deconstructive in that they suggest that apparently self-evident judgments of inherent quality are socially constructed. In my own work on Beethoven’s reputation, for example (DeNora 1995), I was interested in the interaction between Beethoven’s reputation and the organizational culture and practices that allowed Beethoven to be increasingly perceived (and behave) as Vienna’s “greatest” composer. This project was by no means posed in contradiction to the idea of musical value (as some musicological critics believed, e.g., Rosen 1996, DeNora and Rosen 1997), but was rather concerned with two main sociological issues. The first was how, to be a social fact, value of any kind needs to be recognized socially. Unlike gravity or the sound barrier, artistic value is an institutional fact, not a natural one; hence, if it is to be valued, music must be socially recognized and institutionalized as valuable, particularly when it is perceived as violating the norms and conventions that characterize a musical field—when in other words, as with Beethoven, its acceptance constitutes a significant reorientation of taste. (The point is not to presume there is anything automatic about these recognition processes, but to explore them to see how they took shape.) The second issue concerned how the musical field was in flux during Beethoven’s first decade of operation in Vienna, being increasingly transformed in ways that were conducive to the perception of Beethoven’s “greatness”: somewhat like a financier, Beethoven gathered increasing means with which to launch increasingly ambitious aesthetic ventures, while simultaneously augmenting his power within the evaluative terrain of that field. In short, I tried to document the fundamentally practical aspects of how one can emerge as a socially recognized “genius,” so highlighting the way in which genius, as a social fact, emerges from a particular configuration of evaluative criteria, aesthetic orientation and convention, social acts, discourses, and material culture. The study thus focused on the complex interaction between what Beethoven did, what he could do, and how he was perceived. Methodologically, the work began with an investigation of three interrelated factors: the organizational context of music patronage as Beethoven entered it in 1792, his social network as it expanded over time, and his social situation as compared to that of some of his competitors. From there, I adapted methods of ethnographic observation for use on historical data, focusing on agents and actions within this musical field—and specifically on the entrepreneurial activities of Beethoven and his patrons as they presented him in contexts that would flatter his talent. Here the data were letters, other accounts, and contemporary descriptions of the ways in which Beethoven was presented to the public and quasi-public worlds of Viennese musical culture. These were, as I have said, highly pragmatic activities accomplished by Beethoven and his supporters, and they included such things as Beethoven’s own negotiations with the editor of the Allgemeine musikalische Zeitung (a leading music periodical of the day) and his interventions in the world of piano technology. While the study’s aims were ultimately sociological rather than musicological — to theorize, via a case study, key issues concerned with the politics of identity—the book also sought to highlight the contingent nature of the writing of history: in relation to music scholarship, this can be understood as a move away from hagiography and toward an ethnomusicological perspective as applied to the canon. This line of enquiry has been pursued by sociologists in relation to other art

Musical Practice and Social Structure

forms — for example, Heinich’s (1996) study of van Gogh’s posthumous reputation. It has also been pursued as a collaborative project between a musicologist ( J.-M. Fauquet) and a sociologist (Antoine Hennion), in a recent study which argues that the present-day understanding of J. S. Bach is a particular “use” of the composer within a social context (Fauquet and Hennion 2000). By this they mean that the way in which Bach is configured—his value and the ethos for which he is said to stand —represents a form of cultural “work”: it is a tool with which social realities are established and elaborated. The nineteenth-century discovery of Bach and his installation as the “father of music,” Fauquet and Hennion argue, were also a means of configuring the present; Bach’s presence was a resource for articulating the meaning of what it was to be “modern” (Hennion and Fauquet 2001). In this case the empirical strategy was anthropological: Fauquet and Hennion followed various musical (and musicological) actors as they appropriated Bach and so simultaneously produced “Bach” and themselves, defining their own identities in relation to music and, through music, to the social world.

Musical Taste, Consumption and Identity By definition, sociological studies of musical value and its articulation address the matter of how music is appropriated and how music consumption is linked to status definition. This program is implicit in the work discussed in the previous section, and is in turn buttressed by quantitative studies of arts consumption that document links between musical taste and socioeconomic position. In a review of the 1982 national Survey of Public Participation in the Arts, collected for the National Endowment of the Arts by the U.S. Census Bureau, Peterson and Simkus (1992) examined arts participation in relation to occupational group (as a measure of social status). Their aim was to test the notion, as elaborated in Bourdieu (1984), that there is a direct correlation between high social status and the consumption of “high” cultural goods. To do this they considered the case of musical taste, examining items from the survey that addressed musical genre preferences, and attendance at types of music performances. They concluded that, in recent years, perhaps particularly in the U.S., the traditional highbrow / lowbrow division of musical taste has been transformed in favor of an omnivore-univore model. The latter model suggests that individuals with high occupational standing are omnivore-type music consumers: they attend and consume a variety of musical genres. Members of lower status occupational groups, by contrast, exhibit more restricted taste preferences (univores) and are also more likely to defend those preferences vehemently. Quantitative modes of analysis have an important place within the sociology of music. Representative sampling techniques permit reliable and generalizable portraits of populations, which in turn permit the testing of hypotheses — in this case, concerning cultural consumption and social exclusion. But, as with all methods, quantitative techniques pose limits, even when practiced at their best. Peterson’s and Simkus’s work (1992), for example, points directly to questions concerning music and the construction of self- and group-identity; most of these concern the socialpsychological and cultural aspects of musical consumption and practice — music’s

45

46

EMPIRICAL MUSICOLOGY

link, for example, with the social identities of its consumers, its role within sub- and small-group cultures, and its social uses within music-consuming worlds. And nowhere is this tradition better illustrated than in the pioneering work of Paul Willis (1978; see also Frith 1981), with its ethnographically oriented work on the sociology of popular music consumption. Willis was concerned with how, in and through musical practice, through situated consumption of (and talk about) music, musical structures could be seen to have social-organizational properties and capacities. Methodologically, his study drew upon participant-observation techniques (see chapter 2, this volume). The great advantage of this kind of ethnographic observation is its ability to illuminate the nondiscursive dimensions of action (such as emotions and embodiment)—the very dimensions overlooked by survey questionnaires and quasi-formal interview techniques (and also the dimensions of human existence most closely associated with music and musical response). Because of its aims, ethnography is conducted in real time and on the social territories germane to the research subjects themselves. If the aim of one’s research is to understand how music functions, for example, how it inscribes social relations, or how it may serve to inculcate modes of agency within social settings (questions that hark back to Adorno’s concerns), then the advantages of this approach more than outweigh its practical disadvantages (i.e., that it is labor and time intensive, focused on a particular milieu, and not conducive to generalization). In particular, ethnography’s advantage lies in its holistic focus and the emphasis on the emergent and negotiated character of meaning within social settings (Hammersley and Atkinson 1995). Ethnography, in short, can illuminate music as it functions as a resource for meaning construction and for the structuring and organization of social settings. Describing ethnographic work with two groups of music consumers, the “hippies” and the “bike boys,” Willis made his theoretical and methodological perspective clear in the book’s appendix, where he emphasized the virtues of participant observation and its ability to follow actors in natural environments and situations. When allied with other methods, he argued, it provided a means of understanding members’ practices and meanings while suspending theoretical notions that might otherwise be externally imposed. His study involved “hanging around” with members of a motorbike club in an English city, engaging the men in group discussions (tape recorded) where records were played and discussed, and where conversation took off without prompting by the researcher; in the same way, Willis investigated the hippy scene by visiting three groups at their “pads” and holding similar discussion sessions with them. Through this unobtrusive mode of inquiry, held on the respondents’ normal territory and following their ordinary conversation and action, Willis was able to observe how deeply music was implicated in the life worlds of his informants: compared to those of the hippies, the preferred songs of the bike boys were fast-paced and characterized by strong beats and pulsating rhythms. It is here that we can see the great advance of Willis’s study, particularly in its handling of the “homology” concept. While Willis suggested that the preferred music of each group resonated with or was homologous to his groups’ values and habits of being, his concern was to show how the boys themselves established these connections, how they themselves constructed the links between their preferred forms of music and social life. This point

Musical Practice and Social Structure

bears underlining: the structural similarities between music and social organization documented in Willis’s book were forged through the cultural practices and lay classifications of the group members. And it follows from this that, as Willis (1978: 193) put it, “objects, artifacts and institutions do not, as it were, have a single valency [one could read here also ‘single social significance’]. It is the act of social engagement with a cultural item which activates and brings out particular meanings.” In Willis’s work, then, we can observe a theory of musical meaning as located in the interaction between musical objects and music’s recipients; in this respect, Willis’s work connects with other, more theoretically oriented, perspectives within music sociology that conceptualize musical meaning as the result of an interaction between music’s properties (its mobilization of familiar or “stock” materials, conventions, styles, gestures) and the ways these properties are received and responded to (DeNora 1986; Martin 1995). While emphasizing the social construction of meaning, then, Willis is by no means dismissive of the ways in which music’s specific properties may lend themselves with greater or lesser degrees of fit to particular interpretations and appropriations. In the theoretical appendix to his work (1978: 200–201), he describes how cultural items possess “objective possibilities,” but suggests that The same set of possibilities can encourage or hold different meanings in different ways. They can reflect certain preferred meanings and structures of attitude and feeling. On the other hand, because they relate to something material in the cultural item, something specific, unique and not given from the outside, the “objective possibilities” can also suggest new meanings, or certainly influence and develop given meanings in unexpected directions. This uncertain process is at the heart of the flux from which the generation of culture flows. The scope for the interpretation or influence of the “objective possibilities” of an item is not, however, infinite. They constitute a limiting as well as an enabling structure. It is also true that what has been made of these possibilities historically is a powerful and limiting influence on what is taken from them currently. Willis’s work demonstrates that if our aim is to understand music’s social significance and dynamic relationship to social structure, we need to move beyond an exclusive concern with “the music itself” and investigate the processes of its reception and use. This line of thinking has been developed by sociologists of other cultural media: literature (Griswold 1986), television (Moores 1990), and theatre (Tota 1997). Across these studies, attention has been devoted to the more general fabrication of meaning and aesthetic response (including nonverbal response) through interaction with cultural texts, in ways that are directly linked to identity and world construction. The observation that agents attach connotations to things, and orient to things on the basis of perceived meanings, is a basic tenet of interpretivist sociology. But its implications for theorizing the nexus between aesthetic materials and society are profound. It signals a shift in focus from aesthetic objects and their content to the cultural practices in and through which aesthetic materials are appropriated and used to produce social life. And with this shift, we have moved from the cultural constructionism characteristic of recent trends in musicology (as described at the outset of this chapter) to the interactionist constructionism of sociology proper.

47

48

EMPIRICAL MUSICOLOGY

In Willis’s study, “the boys” are seen as interpretatively active; their group values are “almost literally seen in the qualities of their preferred music” (p. 63). The focus is directed at the question of how particular actors make connections or, as Stuart Hall later put it, “articulations” (1980, 1986) between music and social formations. This approach grounds the concept of homology by focusing on the way in which homologies are created (articulated) and experienced, rather than seeing them as inherent in the relationship between pre-given musical texts and pre-given social contextual factors. The further development of this perspective is, arguably, one of sociology’s greatest contributions to the understanding of culture, insofar as it has provided concepts and descriptions of how aesthetic materials come to have, as Willis puts it, social “valency” in and through their circumstances of use. And to see how this valency is produced, ethnographic methods of observation are required— methods that, through their very time-intensity, allow the researcher to observe articulations in the making, in real time and within naturally occurring situations. In the two decades that have followed the publication of Willis’s book, the field of audience and reception studies has advanced considerably. But the early interactionist promise of the classic works of Willis, Frith, and Hall is too often neglected in favor of a preoccupation with the specifics of one or another interpretation of a particular cultural work. The great contribution of these writers was their focus on what the appropriation of cultural materials achieves in action, what culture “does” for its consumers within the contexts of their lives and how these processes can be observed ethnographically. Thus one of the most striking (and usually underplayed) aspects of Willis’s study is its conception of music as an active ingredient of social formation. The bike boys’ preferred music didn’t leave its recipients “just sit[ting] there moping all night” (1978: 69): it invited, perhaps incited, movement. As one of the boys put it, “if you hear a fast record you’ve got to get up and do something, I think. If you can’t dance any more, or if the dance is over, you’ve just got to go for a burn-up” (1978: 73). Willis’s work was pioneering in its demonstration of how music does much more than “depict” or embody values. It portrayed music as active and dynamic, as constitutive not merely of “values” but of trajectories and styles of conduct in real time. It reminded us of how we do things to music and we do things with music, dance and ride in the case of the bike boys, but beyond this, work, eat, fall asleep, dance, romance, daydream, exercise, celebrate, protest, worship, meditate, and procreate with music playing. As one of Willis’s informants put it, “you can hear the beat in your head, don’t you . . . you go with the beat, don’t you?” (1978: 72). As it is used, both as it plays in real time and as it is replayed in memory, music also serves to organize its users’ actions and experiences.

Musically Inscribed Music Consumers Studies such as those conducted by Willis and Frith during the 1970s have proposed that, for the sociology of music, one of the most fruitful analytic strategies is the focus on musical practice. In recent years, the ethnographic focus on musical consumption and musical practice has embraced sociological questions concerned with

Musical Practice and Social Structure

collective behavior and social institutions, as well as questions concerned with the emotions and embodiment. A common thread running through nearly all of the new sociology of music is the concern with music as a resource for social action and for agency broadly conceived. Within social movement theory, for example, music has been conceptualized as providing “exemplars” or models within which social action and movement activity is constructed and deployed (Eyerman and Jamieson 1997). In this respect, music provides, as earlier ethnographers of musical subcultures suggested, a resource for articulating meanings that apply beyond the sphere of music itself. Following actors ethnographically, as they explain themselves in terms that make reference to music, and as they compare themselves or their action styles to musical works, shows how music may actually “get into” action in specific ways, how it functions as an analogue or paradigm for action and cognition. This perspective develops the assumption outlined by Willis’ work on the bike boys, that music provides homologous resources for imagination and conduct. This is saying much more than that there are parallels between music and social forms; it is saying how such parallels are drawn and acted upon—how, as Middleton puts it in his description of Levi-Strauss, music comes to offer “a means of thinking relationships . . . as this note is to that . . . so X is to Y” (Middleton 1990:223). Examining music as it provides media for building social and conceptual relations both extends and operationalizes Attali’s (1985) vision of music’s “annunciatory vocation,” its ability to presage social structural developments. It does this by shifting sociomusical interrogation away from a focus on “reciprocal interactions,” homologies or structural similarities between “music” and “society” (as if these were two distinct realms): instead, it directs focus to the interactive relationship between music and social activity, music and interpretation. This is a pragmatic approach to the topic of musical meaning, one that sidesteps the text/context dichotomy (and the idea of the musical object) in favor of a notion of music as it is drawn into and becomes a resource for action, feeling, and thought. This focus on music as resource has recently been applied to the question of subjectivity and its cultural or social construction (Hennion 1993, Gomart and Hennion 1999, Bull 2000, DeNora 2000, DeNora 2001). Here music is portrayed as a resource for the production and self-production of emotional stances, styles, and states in daily life, and for the remembering of emotional states. The predominant methodological strategy within this work has been the ethnographic interview, designed to uncover, in the first instance, musical practices of the kind that often pass unnoted by respondents—for example, whether they listen to certain types of music in particular circumstances but not in others, or whether they ever choose to listen to works to realign their emotional or energy state. Although this work clearly connects with research in social psychology (Sloboda 1992, Sloboda 2000) and ethnomusicology (Crafts, Cavicci, and Keil 1993), it also indicates an explicitly sociological focus on self-regulatory strategies in particular social contexts. It reveals some of the ways that individuals and groups engage in emotional management and in selfproduction across a range of circumstances. Concurrent with sociological studies of music and emotion management, there has been a renewed interest in music’s effect on and relation to the body. This focus moves beyond the interest, within musicology, in body imagery (Leppert 1993, Walser

49

50

EMPIRICAL MUSICOLOGY

1993) to a concern with bodily praxis and bodily phenomena. In this respect, it connects with recent work by music scholars on the topic of performance “ergonomics” and the socially communicative body in performance (viewing music performance as just one type of social performance [Clarke and Davidson 1998]), and with work on the body as implied and afforded by musical form (McNeill 1995). In keeping with sociology’s emphasis on the situated construction of musical response, new research by sociologists on how music may be understood to mediate corporeal states (such as energy, coordination, entrainment, and bodily self-awareness) downplays a conception of music as stimulus and highlights instead music’s capacity to “afford” — to provide resources for and to enable forms of corporeal organization and states of being. In its focus on music’s connection with modes of being and modes of attending to the social environment, this work connects with Schutz’s classic emphasis on the phenomenological dimension of music making (Schutz 1964). These issues can be illustrated through a study of my own on the role played by music within fitness classes (DeNora 2000: 88–102). The research site — the aerobics class —was chosen because, given the music-led, choreographed character of aerobic exercise, it provided a venue in which music’s role in relation to bodily phenomena (energy, stamina, pain perception, coordination, and motivation) was critical, and where it could, potentially, be observed. The central aim of the research was to illuminate the way in which music structures physical activity and the subjective dimension of that activity. To that end, the study was designed to observe what was conceptualized as “human-music interaction”—the points where music came to serve as an organizing device for bodily activity. It drew together a range of methodological strategies, employed in the following order with overlap between the different types of data collection: participant observation of fitness sessions (primarily “hi/lo” aerobics; the research was undertaken by Sophie Belcher, the extremely fit research assistant); in-depth interviews with music producers; in-depth interviews with class instructors and class members; and quick questionnaires, administered to class members. Given the aims and subject matter of the study, participant observation was a critical investigative technique. As with most embodied practices, there are many things about aerobics that one can only know about by doing. Being physically stretched, for example, experiencing “the burn,” sweating and tuning into the rhythm of a session, feeling at the point of fatigue and then re-energized when the music changes, wanting to move with gusto to the musical pulse—these are all experiential matters. The first form of data in this study thus consisted of the (junior) researcher’s own experience of exercising to music, her “knowing-by-doing.” The second form of data was the record provided by the videotapes of each session. These enabled the researcher not only to recall the embodied experience of class sessions, but also to see and freeze otherwise fleeting and subconscious moments of class experience, to play them back and so enable reflection upon what it was about the music that enabled or constrained forms of physical activity. This reflection was facilitated through conversations (in-depth interviews) with the senior researcher (myself ), such that the research assistant was, simultaneously, researcher and key informant in the study. These conversations (analytically oriented debriefing sessions) in turn generated hypotheses and ideas for further observation. Key among these was the strategy of ex-

Musical Practice and Social Structure

amining “breakdowns” in sessions and of comparing “good” sessions with “bad” ones, that is, sessions characterized by a high degree of aerobic order with sessions where such things as fatigue, lack of coordination, and boredom occurred. The third form of data came from interviews with professional aerobics music providers. Here the focus was on what these providers said about music’s features and their usefulness for exercise. The data from these interviews were compared with what class participants themselves thought about particular musical numbers and passages, types of movement, and their associated motivational states. The research highlighted ways in which specific musical devices were enabling or constraining at certain points in the aerobic session. The key point was that some of these devices were effective for some aspects of the exercise session but not for others, and this finding helped to highlight music’s structuring properties in relation to the body and embodied activity. From there, the key research question became why certain features were effective at certain stages of the aerobic process. Analysis of all the data, and particularly the videotapes, suggested that music could be seen to work with and for the body (and against the body) by profiling bodily movement, by entraining movement, and by modeling and enabling the adoption of motivational stances (and energies) appropriate to different segments of the session. So, for example, slower-paced, more “lyrical” formats were useful for the stretching movements of the warmup, while music with a highly prominent beat and powerful orchestration (e.g., lower brass tutti) was useful during the core of the session characterized by a vigorous movement style. The study concluded that music could serve as a “prosthetic technology” (Ehn 1988: 399) of the body, a device that has the capacity to extend and restructure bodily phenomena, including embodied states such as emotion and motivation. This capacity is by no means confined to the totalizing environment of the exercise session, but can be perceived across a range of settings in daily life—in the workplace, within organizations (Lanza 1995), and in commercial environments such as restaurants and shops. Indeed, this study was followed by an ethnography of music in retail outlets, with an explicit focus on these outlets’ attempts to configure modes of agency (here understood as predispositions for and styles of action or subjectivity) by configuring the sonic environment (DeNora 2000, chapter 5). Overall, we were interested in how shops used music to target preferred types of consumers, and to structure the temporal and other aspects of the environment; and we were also interested in how shoppers interacted with music in-store—for instance whether they noticed it and, if so, what they thought about it. As with the aerobics work, our own autobiographical experiences in relation to in-store music were used as a basis for generating interview questions and as a ground against which to analyze the in-store conduct of other shoppers. To these ends, and with the permission of the stores in question, the research assistant and I posed as shoppers to observe the scene in-store, and in particular to take note of (and compare) in-store ambience and the conduct of other shoppers. With tape recorders unobtrusively held in our hands (they pass as personal stereos) and clip-on microphones on our coat collars, we simultaneously recorded the instore soundtrack and our various observations about the conduct of other shoppers (such as whether shoppers showed signs of engaging with the music by singing along,

51

52

EMPIRICAL MUSICOLOGY

snapping their fingers, or making dance movements). We combined this with semistructured interviews with shop managers and staff about their music policies and how the music seemed to work in-store, as well as exit interviews with consumers as they left the shops (we did not have permission to speak to shoppers in-store). In addition, as a pilot study, we followed volunteer shoppers whom we “wired for sound”: we asked them simply to “think out loud” as they moved through the shop, commenting on anything that came to mind and anything they might notice about the music in particular. Simultaneously, we shadowed these shoppers, one-on-one, recording our own observations about their behavior (e.g., “she is looking at an orange jumper now”). The two tapes could be synchronized precisely because they shared the same musical soundtrack, and this enabled us to overlap the two transcripts.

Conclusion I have sought, in this chapter, to unpack the sociology of music’s toolkit, and to feature, in particular, the tools designed for the exploration of musical practice; some of these tools are new, and their utility over the long term is yet to be determined. I have also sought to highlight how the sociology of music currently elaborates a particular version of constructionism, one that takes as its object of analysis music as it is made, used, and responded to within specific contexts and settings. Sociologists of music have drawn upon a range of empirical strategies for the collection and analysis of music’s social role, from analyses of networks and the impact of associations and information exchange on music-stylistic choices, to comparative analyses of institutional and organizational structures of music making and their influence on musical works and producers. Survey methods have been used for mapping musical participation and taste, and in-depth interviews for exploring reception issues; ethnographic methods have been adopted for the examination of music as it is involved and mobilized in culture creation, group culture, and the noncognitive dimensions of social being and social life. Over the past three decades, the sociology of music has shifted from its status as a somewhat abstract endeavor located on the margins of sociology, to a grounded and empirically oriented mode of enquiry directed to many of sociology’s core concerns—social structure, consciousness, and social difference and division. In undergoing this change, the sociology of music has not only been empowered within sociology as a whole, but has simultaneously retooled in ways that are significant for musicology, as that field develops toward an understanding of music as a fundamentally social enterprise.

References Adorno, T. W. (1973). Philosophy of Modern Music. New York: Seabury. Adorno, T. W. (1976). Introduction to the Sociology of Music. New York: Seabury. Adorno, T. W. (1990). “On popular music,” in S. Frith and A. Goodwin (eds.), On Record: Rock, Pop and the Written Word. London: Routledge, 301–314.

Musical Practice and Social Structure Attali, J. (1985). Noise: Toward a Political Economy of Sound. Manchester: Manchester University Press. Becker, H. S. (1963). Outsiders. New York: Free Press. Becker, H. S. (1982). Art Worlds. Berkeley: University of California Press. Becker, H. S. (1989). “Ethnomusicology and sociology: A letter to Charles Seeger.” Ethnomusicology 33: 275–285. Berger, B. (1996). An Essay on Culture. Berkeley: University of California Press. Bijker, W. E., T. P. Hughes, and T. Pinch (eds., 1987). The Social Construction of Technological Systems. Cambridge, Mass.: MIT Press. Bourdieu, P. (1984). Distinction: A Social Critique of the Judgement of Taste, Cambridge, UK: Polity Press. Bull, M. (2000). Sounding out the City. Oxford: Berg. Cerulo, K. A. (1984). “Social disruption and its effects on music—an empirical analysis.” Social Forces 62: 885–904. Citron, M. (1993). Gender and the Musical Canon. Cambridge: Cambridge University Press. Clarke, E., and Davidson, J. (1998). “The body in performance,” in W. Thomas (ed.), Composition—Performance—Reception. Studies in the Creative Process in Music. Aldershot: Ashgate Press, 74–92. Cook, N. (1990). Music, Imagination and Culture. Oxford: Clarendon Press. Crafts, S., Calvacci, D., and. Keil, C. (1993). My Music. Hanover, N.H.: Wesleyan University Press. DeNora, T. (1986). “How is extra-musical meaning possible? Music as a place and space for work.” Sociological Theory 4: 84–94. DeNora, T. (1995). Beethoven and the Construction of Genius. Berkeley: University of California Press. DeNora, T. (2000). Music in Everyday Life. Cambridge: Cambridge University Press. DeNora, T. (2001). “Aesthetic agency and musical practice: New directions in the sociology of music and emotion,” in P. Juslin and J. Sloboda (eds.), Music and the Emotions. Oxford: Oxford University Press, 161–180. DeNora, T., and Rosen C. (1997). “‘Beethoven’s genius’: An exchange.” New York Review of Books XIV/6 (April 10, 1997): 66–68. DiMaggio, P. (1982). “Cultural entrepreneurship in nineteenth-century Boston: The creation of an organizational base for high culture in America.” Media, Culture and Society 4: 35–50, 303–322. Edström, O. (1997). “Fr-a-g-me-n-ts: A discussion of the position of critical ethnomusicology in contemporary musicology.” Svensk Tidskrift för Musikforskning, 79: 9–68. Ehn, P. (1988). Work-oriented Design of Computer Artifacts. Stockholm: Arbetslivscentrum. Elias, N. (1993). Mozart: Portrait of a Genius. Cambridge, UK: Polity Press. Eyerman, R., and Jamieson, A. (1997). Music and Social Movements. Cambridge: Cambridge University Press. Fauquet, J-M., and Hennion, A. (2000). La grandeur de Bach: L’amour de la musique en France au XIX siecle. Paris: Fayard. Frith, S. (1981). Sound Effects: Youth, Leisure, and the Politics of Rock ’n’ Roll. New York: Pantheon. Frith, S. (1996). Performing Rites: Evaluating Popular Music. New York: Oxford University Press. Goehr, L. (1992). The Imaginary Museum of Musical Works: An Essay in the Philosophy of Music. Oxford: Clarendon Press.

53

54

EMPIRICAL MUSICOLOGY

Gomart, E., and Hennion, A. (1999). “A sociology of attachment: Music amateurs, drug users,” in J. Law and J. Hazzart (eds.), Actor Network Theory and After. Oxford: Blackwell, 220–247. Green, L. (1997). Music, Gender and Education. Cambridge: Cambridge University Press. Griswold, W. (1986). “The fabrication of meaning.” American Journal of Sociology 92: 1077–1117. Hall, S. (1986). “On postmodernism and articulation: An interview with Stuart Hall.” Journal of Communication Inquiry 10: 45–60. Hall, S. (1980). “Recent developments in theories of language and ideology: A critical note,” in S. Hall, D. Hobson, A. Lowe, and P. Willis (eds.), Culture, Media, Language: Working Papers in Cultural Studies 1972–79. London: Hutchinson, 157–162. Hammersley, M., and Atkinson, P. (1995). Ethnography: Principles in Practice (rev. ed.). London: Routledge. Heinich, N. (1996). The Glory of van Gogh: An Anthropology of Admiration. Princeton, N.J.: Princeton University Press. Hennion, A. (1989). “An intermediary between production and consumption: The production of popular music.” Science, Technology and Human Values 14: 400–424. Hennion, A. (1993). La Passion musicale. Paris: Metaille (forthcoming as Music as Mediation, Manchester: Manchester University Press). Hennion, A. (1995). “The history of art—lessons in mediation.” French Journal of Communication 3: 233–262. Hennion, A. (1997). “Baroque and rock: Music, mediations and musical taste.” Poetics 24: 415–435. Hennion, A., and Fauquet, J.-M. (2001). “Authority as performance: The love for Bach in nineteenth-century France.” Poetics 29: 75–88. Johnson, J. (1995). Listening in Paris: A Cultural History. Berkeley: University of California Press. Juslin, P., and Sloboda, J. (eds., 2001). Music and the Emotions. Oxford: Oxford University Press. Kingsbury, H. (1991). “Sociological factors in musical poetics.” Ethnomusicology 35: 195–219. Knorr-Cetina, K. (1981). The Manufacture of Knowledge: An Essay on the Constructivist and Contextual Nature of Science. Oxford: Pergamon Press. Lanza, J. (1995). Elevator Music: A Surreal History of Muzak, Easy Listening and Other Moodsong. London: Quartet Books. Latour, B. (1987). Science in Action. Cambridge, Mass.: Harvard University Press. Latour, B., and Woolgar, S. (1986). Laboratory Life: The Construction of Scientific Facts. Princeton, N.J.: Princeton University Press. Leppert, R. (1993). The Sight of Sound: Music, Representation, and the History of the Body. Berkeley: University of California Press. Leppert, R., and McClary, S. (eds., 1987). Music and Society. Cambridge: Cambridge University Press. Lomax, A. (1968). Folk Song Style and Culture. Washington, D.C.: American Association for the Advancement of Science. Maisonneuve, S. (2001). “Between history and commodity: The production of a musical patrimony through the record in the 1920–1930s.” Poetics 29: 89–108. Martin, P. (1995). Sounds and Society: Themes in the Sociology of Music. Manchester: Manchester University Press. Martin, P. (2000). “Music and the sociological gaze.” Svensk Tidskrift för Musikforskning 82: 41–56. McClary, S. (1991). Feminine Endings. Minneapolis: University of Minnesota Press.

Musical Practice and Social Structure McNeill, W. H. (1995). Keeping Together in Time: Dance and Drill in Human History. Cambridge, Mass.: Harvard University Press. Middleton, R. (1990). Studying Popular Music. Milton Keynes, UK: Open University Press. Moores, S. (1990). Interpreting Audiences. London: Sage. Negus, K. (1992). Producing Pop: Culture and Conflict in the Popular Music Industry. London and New York: Edward Arnold. North, A. and Hargreaves, D. (1997). “Music and consumer behaviour,” in D. Hargreaves and A. North (eds.), The Social Psychology of Music. Oxford: Oxford University Press, 268–289. O’Neill, S. (1997). “Gender and music,” in D. Hargreaves and A. North (eds.), The Social Psychology of Music. Oxford: Oxford University Press, 46–66. Pasler, J. (forthcoming). Useful Music, or Why Music Mattered in Third Republic France. Berkeley: University of California Press. Peterson, R. (ed., 1976). The Production of Culture. Los Angeles: Sage. Peterson, R., and Berger, D. (1990). “Cycles in symbol production: The case of popular music,” in S. Frith and A. Goodwin (eds.), On Record: Rock, Pop and the Written Word. London: Routledge, 140–159. Peterson, R., and Simkus, A. (1992). “How musical tastes mark occupational status groups,” in M. Lamont and M. Fournier (eds.), Cultivating Differences: Symbolic Boundaries and the Making of Inequality. Chicago: University of Chicago Press, 152–186. Pinch, T., and Trocca, F. (forthcoming). Analog Days. The Invention and Impact of the Moog Synthesizer Cambridge, Mass.: Harvard University Press. Randel, D. (1992). “The canon in the musicological toolbox.” In K. Bergeron and P. Bohlman (eds.), Disciplining Music: Musicology and its Canons. Chicago: University of Chicago Press, 10–23. Rosen, C. (1996). “Did Beethoven have all the luck? Beethoven and the construction of Genius.” New York Review of Books XLIII/18 (November 14, 1996): 57–63. Schutz, A. (1964). “Making music together: A study in social relationship,” in A. Brodersen (ed.), Alfred Schutz: Collected Papers (vol. 2). The Hague: Martinus Nijhoff, 159–78. Shepherd, J. (1991). Music as Social Text. Cambridge, UK: Polity Press. Shepherd, J., and Wicke, P. (1997). Music and Cultural Theory. Cambridge, UK: Polity Press. Sloboda, J. (1992). “Empirical studies of emotional response to music,” in M. R. Jones and S. Holleran (eds.), Cognitive Bases of Musical Communication. Washington, D.C.: American Psychological Association, 33–46. Sloboda, J. (2000). “Everyday uses of music listening,” in S. W. Yi (ed.), Music, Mind and Science. Seoul: Western Music Research Institute, 354–369. Sterne, J. (1997). “Sounds like the mall of America.” Ethnomusicology 41: 14–50. Steward, S., and Garratt, S. (1984). Signed, Sealed and Delivered. True Life Stories of Women in Pop. London: Pluto Press. Subotnik, R. (1976). “Adorno’s diagnosis of Beethoven’s late style: Early symptoms of a fatal condition.” Journal of the American Musicological Society 29: 242–275. Subotnik, R. (1978). “The historical structure: Adorno’s ‘French’ model for the criticism of nineteenth-century music.” 19th-Century Music 2: 36–60. Subotnik, R. (1983). “The role of ideology in the study of Western music.” Journal of Musicology 2: 1–12. Subotnik, R. (1990). Developing Variations: Style and Ideology in Western Music. Minneapolis: University of Minnesota Press.

55

56

EMPIRICAL MUSICOLOGY

Tota, A. (1997). Ethnografia dell’ arte. Rome: Logica University Press. Walser, R. (1993). Running with the Devil: Power, Gender and Madness in Heavy Metal Music. Hanover, N.H.: Wesleyan University Press. Weber, M. (1958). The Rational and Social Foundations of Music. Carbondale: Southern Illinois University Press. Weber, W. (1978). Music and the Middle Class. London: Schocken. Weber, W. (1991). The Rise of Musical Classics in Eighteenth-Century England: A Study in Ritual, Canon, and Ideology. Oxford: Clarendon Press. Willis, P. (1978). Profane Culture. London: Routledge. Witkin, R. (1998). Adorno on Music. London: Routledge. Wolff, J. (1981). The Production of Culture. Oxford: Blackwells. Zolberg, V. (1990). Constructing a Sociology of the Arts. Cambridge: Cambridge University Press.

CHAPTER 4

Music as Social Behavior Jane W. Davidson Introduction In the vast majority of music-making contexts, the real or implied presence of others means that at some level social communication or interaction takes place: singing a lullaby, a work song, a hunting song, or a school song; chanting as the member of a football crowd; participating as either a musical performer or a spectator in a symphony orchestra concert or at a Hindu wedding. In fact, individual practice is one of the rare musical occasions when there is no involvement with a co-performer or spectator, but even here there is generally a social goal: the preparation of a performance. Recordings might seem to be another exception, but the social element is still implied: there is a need to communicate the musical content to someone else, even if for the duration of the recording the audience is imaginary. Music is a social act, but investigating how social behaviors function in different musical contexts, and what significance they have, is a very recent research interest in psychological approaches to Western music (for an overview see Hargreaves and North 1999). The delayed development of social psychological research seems to be the result of a largely reductionist approach to music which has tried to understand it in terms of its structural elements: melody, harmony, rhythm, and so forth. But, as general interest in issues related to attitudes and beliefs, and individual and group behavior, has grown, so too has the interest in music as a social-behavioral phenomenon. Within the psychology of music, Farnsworth (1954) was one of the first to exhibit an explicit interest in such issues, arguing that it was not sufficient to look at how a song functioned musically; rather it was important to know how the performing context operated, and how it affected both performer and audience. Anecdotal accounts can go some way toward describing social behaviors, but the motivation behind the work of researchers such as Farnsworth was to undertake more systematic investigations. They wanted to generalize from their observations, measuring the frequencies of musical behaviors and the interrelationships of these behaviors within and across individuals. Thus they adopted quantitative research designs employing statistical techniques in the analysis of data. By the 1970s and 1980s, however, experimental studies were increasingly complemented by work influenced by the writings of theorists like Harré (1979, 1992a, 1992b), who demonstrated that controlled manipulation under experimental conditions was not always an appropriate methodology when looking at beliefs and behaviors. Out of these kinds of theoretical discussion emerged New Paradigm Research, which adopted qualitative research techniques such as in-depth semistructured interviews and par57

58

EMPIRICAL MUSICOLOGY

ticipant observation in an attempt to capture the subjectivity of individual experience (Smith, Harré, and van Langenhove 1995.) As a consequence, there is now a much more diverse palette of research techniques available for investigating music as a form of social behavior than there was at the time of Farnsworth’s work. At one extreme, there is the tradition based on the reductionist approach which manipulates conditions so that specific variables can be examined. An example of this is Burland and Davidson’s (2001) study of the effects of different social groupings (single sex, mixed sex, friends, strangers, and classmates) on the quality of musical compositions, exploring the effects of different social combinations on problem-solving in a musical composition task; without the controlled manipulation of the independent variable (the social group in which the children were working), the study would not have been able to demonstrate the impact of social processes on the dependent variable (the musical composition).1 Formal questionnaires, surveys, and tests have also been used. For instance, Finnäs (1987) wanted to explore the musical preferences of young people; in order to do this, he designed an experiment in which participants listened to and then rated excerpts on a preference scale — a typical quantitative design that is then subjected to statistical analysis. Since Finnäs believed that most participants of this age group would prefer rock over other forms of music, he designed the experiment in such a way that participants always had to compare a test excerpt (a piece of classical music, for instance) with a rock item, so that the relative preference for an item in relation to the best liked piece could be determined. Other investigations have combined qualitative and quantitative methods. Davidson and Scutt (1999) used a quantitative measure of musical achievement (a musical examination grade) as an objective point around which to assess the role of students’, parents’, and teachers’ beliefs and practices in preparing for these examinations, collected by means of qualitative interviews. By contrast, work of an entirely qualitative nature, like that of Kanellopoulos (1999), demonstrates the New Paradigm approach in which the researcher plays an active role in the generation and interpretation of the data.2 Kanellopoulos wanted to know how a class of children worked on free improvisation. The analysis emerged from examining the content of conversations Kanellopoulos had with the class of children, semistructured interviews, and tape recordings of the musical and social interactions during the improvisation process. Kanellopoulos also became a co-improviser, attempting to enter the children’s world by participating in the creative process alongside them. The data revealed the critical role of social context, with classmates and the teacher assisting in the development of musical improvisation skills. In summary, a broad spectrum of research approaches has developed, each approach serving a slightly different purpose. In Burland and Davidson’s study, only a short period of time was available in which to explore the impact of social group on the ability to compose, and so the most effective approach was an experiment that deliberately formed groups and so manipulated the social context in the study. For Finnäs, rating scales produced easily quantifiable data in response to a straightforward question: what are young people’s musical preferences? Davidson and Scutt used data of a more objective nature and compared them with reports of thoughts, feelings, and practices in the preparation, execution, and follow – up periods to a

Music as Social Behavior

public examination; these longitudinal qualitative data permitted emergent social relationships and key themes to be described and explored over a six-month period. Finally, Kanellopoulos undertook qualitative interviews similar to those of Davidson and Scutt, but involved himself fully in the process, reflecting on his own input as teacher and guide to the children. The aim of this chapter is not only to illustrate the breadth of social-behavioral research, but also to look in detail at how music has been analyzed as a form of social behavior, and to pinpoint why particular methods of analysis might be appropriate to specific research questions. It will not be possible to provide an overview of all the forms of social musical behavior, nor to explore the vast range of strategies that might be employed to examine research questions. Instead, the focus will be on a number of specific areas from within the Western art tradition: the music learning environment; social influences on musical behaviors; individual differences in musical behavior; and the social context of musical performance.

Social Factors in Musical Skill Acquisition Research techniques applied to questions of skill acquisition have drawn on a whole spectrum of approaches, from large-scale quantitative questionnaire studies to detailed qualitative case studies. Two contrasting research approaches are considered here: the first involves the collection of biographical data to provide quantitative and generalizable results, while the second considers material focused on the individual and derived from a longitudinal case study.

The Biographical Survey If you want to find out how musical skills are acquired within a social context, musicians’ biographies can be an effective source of information. Biographies are useful because they can trace the factors, social and otherwise, that have led to musical achievement. In terms of existing social-psychological research in music, retrospective biographical accounts have provided the major source of information about key influences on the development of musicians, particularly those of prodigious achievement. Such research has been undertaken in a variety of ways, but principally through interviews in which parents and children talk about their lives and what happened at particular stages of their musical development (e.g., Sloboda and Howe 1991). This is not an optimal research technique, since human memory is notoriously unreliable, but it is often possible to verify retrospective accounts at least in part by asking how, for instance, a child was practicing at a certain time, or looking for data of a more objective nature such as the dates of examinations and grades achieved. Biographical research has clear and established precedents in historical musicology, which can provide a useful resource for subsequent analysis. Lehmann (1997), for instance, carried out an analysis of historical documents relating to famous individuals, cross-referencing a variety of texts to provide both confirmatory evidence and extra detail. While this is a useful method of collecting data, it is also problematic. Historical documents can be unreliable, with accounts from similar periods often

59

60

EMPIRICAL MUSICOLOGY

being summaries of one another and no opportunity available to fill the gaps in what may be an incomplete record. Furthermore, the accuracy of the data can be questionable when it relies on second-hand reports written many years after the actual events. Despite such drawbacks, however, primary source material such as the considerable correspondence between Clara and Robert Schumann (Weissweiler 1984) — or the diaries of Berlioz (Searle 1966), which tell of concert life and the role of various individuals in the development of his musical skills— are important documentary resources, rich in self-reflection and insight. And similar kinds of diary data have been widely used in socio-behavioral enquiries. For instance, in research on pregnancy, Smith (1994) gained all manner of insights into a woman’s sense of changing self from the diary she kept during the course of her pregnancy. But the key to understanding the free accounts in the diaries was the follow-up interview: by asking similar questions of all the women participating in his study, Smith was able to undertake a comparative analysis of both sets of data. The interviews provided information that then permitted all the material to be systematically structured and thematically arranged. This, obviously, is not possible in the case of historical data. In looking at the events underlying musical development, an alternative to the historical biography is a real-time survey over the life-span of many individuals. Researching music over the life span is a tricky business, since in Western culture not everyone learns an instrument, and a very large population survey would be needed to ensure the inclusion of a sufficient number of participants who developed musical skills and then progressed to high levels of accomplishment. Sloboda went some way toward collecting this kind of data by adding a section of questions on music to the Mass-Observation Project based at the University of Sussex, which involves around 500 people in an ongoing recording of their daily lives (Sheridan, Bloome, and Street 1998). In autumn 1997, data on individual involvement with music was sought, with family and other influences being explored in order to find out what proportion of the population engage in music and how this process occurs. Although some elements of these data have been reported by Sloboda in conference proceedings (1998), the main analyses have yet to be undertaken as they are dependent on further data collection. It is generally very difficult to find ways of separating out the many potential factors that may have an effect on an individual or a social group. The next section shows how this difficulty can be addressed through biographical research projects involving empirical methods and comprehensive data collection. There are many possible research techniques, but I discuss the two principal techniques used in behavioral research: the quantitative and the qualitative questionnaire.

Developing a Quantitative Biographical Questionnaire Study In most quantitative, survey-style research, considerable effort is put into making sure that the right sorts of question are asked so that the data collected are appropriate to the particular aims of the investigation. While reviewing the relevant literature can help to give a preliminary indication of potentially valuable lines of enquiry, it is common for a small number of in-depth interviews to be undertaken prior to the main questionnaire, so that the focus of questions can be clarified and devel-

Music as Social Behavior

oped. At this stage, new issues that emerge can be easily integrated into the design. Similarly, there is often a phase of “piloting” the questions and response categories in order to make sure that the data are collected in an appropriate format for analysis, as well as to identify the relevant statistical analysis techniques. Once the questionnaire and analysis technique have been developed, individuals from the sample to be surveyed are often given “trial” interviews to make sure that they can deal with both the types of questions proposed and the interview environment itself. The fine details of preparing questionnaires can be found in Oppenheim’s (1992) text on the topic. However a useful set of examples is provided by a series of studies in which a small-scale biographical study was used as a pilot to develop a major investigation of young musicians with a variety of backgrounds and interests in music (Howe and Sloboda 1991a, Howe and Sloboda 1991b, Sloboda and Howe 1991, Davidson, Howe, and Sloboda 1995 / 1996, Davidson, Howe, Moore, and Sloboda 1996, Davidson, Howe, Moore, and Sloboda 1998). Initially, Sloboda and Howe wanted to find out what social and other factors such as motivation, practice habits and so on were shared by musicians of prodigious talent; the intention was to unravel the issue of how far musical ability is the product of nature or nurture. The researchers chose to survey 42 young musicians of between 8 and 18 years of age, who were all in specialist musical education and had musical accomplishments well beyond those expected of a normal child. This number of children was chosen so that shared factors could be more easily identified and grouped: it would be difficult, for instance, to say that a particular social factor was influential in the development of musical ability if only two out of three interviewees shared it. The 42 children were statistically representative of their group, and were sufficiently numerous to generate statistically analyzable results relating to the behavioral influences on young, talented musicians in Britain. Sloboda and Howe initially worked from the existing research literature on high achievement in order to identify a number of key areas for the interviews: the impact of parents, teachers, siblings, and role models. They also wanted to find out information that was specific to the musical tasks, such as how often the children practiced and what examinations they had taken. Having identified these broad areas of interest, they then constructed an open-ended interview schedule — that is, a schedule that would be delivered in an informal conversational style. In open-ended questioning of this kind, the interviewer needs to know the interview topic areas very well, and be prepared to go along with the flow of a conversation from one topic to the next at the participant’s pace; he or she needs to pursue areas that may emerge spontaneously in the interview, and that may end up providing important insights into the participant’s thoughts about an issue. Despite the open nature of these interviews, however, Sloboda and Howe made sure that they began with general, noninvasive areas of questioning, and then as the interview went on, proceeded to more specific questions, with the ultimate aim of finding out more personal or potentially sensitive information. As Smith, Flowers, and Osborn (1997) have pointed out, this kind of funneling of questions allows for a rapport to be established between interviewer and interviewee prior to tackling questions of a personal nature. After carrying out their interviews with the young musicians in this semistructured manner (one-to-one and tape recorded), Sloboda and Howe developed a

61

62

EMPIRICAL MUSICOLOGY

second schedule for the musicians’ parents, who were all interviewed independently. Each interview lasted approximately 40 minutes and full transcripts of all interviews were made. At this stage, the researchers examined all the data for common themes, and for any corroboration between the parent and child interviews that might validate what had been said; this approach produced more themes and topics than the researchers had initially anticipated, resulting in a rich data set that was highly specific to the individuals interviewed. Once they had grouped the data according to theme, the researchers undertook statistical analyses of the frequency of different behaviors and the correlations between practice and achievement, exploring those factors that contributed most to these children’s musical progress. These statistical analyses (reported in Sloboda and Howe 1991) were used to show that the children they interviewed were significantly supported by social networks of family, friends, and teachers. Although the data were dependent on retrospective memory, and for that reason potentially unreliable, the independent interviews with children and parents allowed for differences in recollections to be accounted for. Moreover, by interviewing children still in education, the interviewers made a deliberate attempt to collect data as near as possible to the time that the relevant events occurred. In the three publications that resulted from their biographical investigation of the 42 children (Howe and Sloboda 1991a, Howe and Sloboda 1991b, Sloboda and Howe 1991), Sloboda and Howe highlighted different elements of the data. The first paper looked for quantifiable results to show trends for the particular group of individuals investigated, while the subsequent two papers dealt much more directly with the data as they were collected, for instance providing extensive quotes to illustrate how children from professional musician families saw music as an integral part of their lives. In a text on research techniques in the social sciences, Robson (1993) notes that some of the best research is that which combines the generalizability of statistical findings—for example, that 73% of those interviewed had supportive parents— with the particularities of individual cases. The study reported above, however, was not only an end in itself, but also provided pilot data for a much larger-scale and more formally structured questionnaire study which combined past and current accounts of musical engagement. Developing and refining the research questions and expanding the population to be surveyed to include children who had given up music as well as those who had music as a second or third hobby, Sloboda and his colleagues developed a formal multiple-choice questionnaire suitable for 45-minute interviews with over 250 children and their parents. Like the initial study, this larger-scale survey involved interviewing all children and parents individually so that comparisons between the data could be made. Besides the more formal retrospective interviews concerning family, sibling, and teacher influences, the children were asked to maintain diaries of their lessons, practicing, and other forms of engagement with music. These diaries produced a more accurate survey of what children were actually doing by reducing the problems of reconstruction and retrospection. One group of children investigated by the researchers (those who had aspirations to become professional musicians but were not receiving a specialist musical education) came from a wide geographical spread, so postal survey techniques were

Music as Social Behavior

adopted in order to collect data from them. Oppenheim (1992) observes that it is not untypical for return rates on postal surveys to be as low as 15%. Such low returns may be a depressing prospect, but an astute researcher can increase potential return rates in a number of ways. First and most obviously, material rewards such as a free CD can be offered to participants, putting them under a certain obligation to respond. Second, the questionnaire should be designed and expressed in such a way that the participants feel that what is being requested of them will be of value to the research. Third, the questionnaire should not be too long or complex. Careful piloting of how questions are constructed and set out on the page is just as important as the content and relevance to the potential respondent: for instance, a questionnaire might begin with easy, general questions and work toward more specific and “difficult” ones (such as those that are of a personal nature or require a lengthy written description as opposed to a one-word response). Sloboda and his colleagues meticulously piloted their postal questionnaire, modifying wordings of questions and response sheets in the light of their face-to-face interviews, and their careful planning resulted in a return rate of around 50 percent. This large-scale study demonstrated that children who regarded musical participation as a second or third hobby alongside sports or other activities, or gave up musical participation completely, had entirely different family, school, and peer circumstances compared with those children who attained high levels of achievement. From this, the researchers were able to gain some insight into the role of environmental influences in the development of musical ability, which in turn enabled them to engage in a much broader theoretical debate concerning the relative roles of nature and nurture in musical development. (These specific debates can be found in Sloboda, Davidson, and Howe 1994a, Sloboda, Davidson, and Howe 1994b, Howe, Davidson, and Sloboda 1998a, Howe, Davidson, and Sloboda 1998b.) In summary, Sloboda and his colleagues aimed their research at discovering general characteristics and overall trends, and they used the semistructured interview primarily as a starting point from which to develop a fully structured questionnaire. By contrast, the next case study uses the semistructured interview as a means of producing a very detailed case study of a single family’s views and behaviors with regard to music.

The Case Study and Interpretative Qualitative Analysis A series of publications by Borthwick and Davidson tackles the issue of how the family, as a social unit, supports children’s musical learning (Borthwick 2000; Borthwick and Davidson 2002; Davidson and Borthwick 2002). Borthwick noted that although researchers like Sloboda and Howe had considered the influence of others in the child’s acquisition of musical skills, they had not looked at interactions within the immediate family and how these might influence musical development. She therefore focused her work on the complexity of the interactions within a specific family: parent – child, parent–parent, and sibling–sibling. Overall, her work demonstrated not only that the elder sibling provided a role model for the younger one, but also that the parents established a strong “script” in relation to their children’s music. The concept of a script was adapted by Borthwick from the work of Byng-Hall (1995), a

63

64

EMPIRICAL MUSICOLOGY

family therapist, who defines a “script” as a set of beliefs and behaviors that regulate the social roles played by each individual within the family. By keeping detailed records of observed and reported behaviors and beliefs over a period of 18 months, Borthwick was able to argue that music scripts exert a powerful influence on children’s musical development and their perception of themselves in relation to other family members. For instance, exploring parent– child coalitions, she discovered that the father seemed to project his own musical identity as a professional pianist onto his eldest daughter, and that the daughter took on the script as part of her selfidentity, believing that she was the inheritor of her father’s abilities. This approach is interesting for two reasons: first, the “script” theory provides a framework within which to interpret the highly subjective material that Borthwick collected from the family; and second, Borthwick took on the role not only of interviewer, but also of observer of, and participant in, the family’s life. Visiting the family every two weeks for interviews, she was able to observe the children’s music making in the home and how it was discussed and supervised by the parents. She was also invited to participate in family events like Easter egg making, so that from early in the research period she faced few of the “outsider” difficulties that an interviewer coming to a family for the first time might encounter. Borthwick’s approach generated a depth of data far in excess of that produced by Sloboda and his colleagues. However it was extremely time consuming, and the data analysis (based on transcripts of interviews, conversations, and a series of diaries that she maintained) was very detailed and labor intensive. Although the use of such material is well established in ethnomusicology (see chapter 2, this volume), it is only recently that interview, discourse, and diary text has been used for the psychological analysis of musical behavior (e.g., Davidson and Smith 1997). Within psychological, sociological, and educational frameworks more generally, however, the 1990s saw an increase in qualitative research of this kind. Techniques such as verbal protocol analysis, discourse analysis, grounded theory, and interpretative phenomenological analysis have emerged (for an overview, see Robson 1993), all of which use small numbers of participants or even single case studies. In all these techniques, the intersubjectivity between the participant and the researcher is acknowledged as part of the research process, requiring that the researcher’s own thoughts and feelings about the participant are discussed and included as a critical part of the analysis. Measures of reliability are not attainable in the same way as in quantitative research (that is to say, by conventional statistical methods), but this problem can be tackled in two different ways. One approach is for the researcher to ask another person with similar analytical skills to examine both the raw data and the analysis in order to evaluate the internal consistency of the researcher’s analysis. An alternative is for the researcher to collect multiple forms of data (e.g., interview and diary data) and to use these different sources to “triangulate” the data—that is, to access the participants’ thought processes from a variety of convergent perspectives. In either case, the analyses are often taken back to the participant, who may be asked whether or not the interpretations are plausible. The research methodology adopted by Borthwick is highly interpretative and has its roots in the belief that individuals construct a sense of reality on the basis of understanding and interacting with a social world (see Gergen and Davis 1985). The

Music as Social Behavior

data are discussed as constructions rather than realities, and as a researcher Borthwick considered reflexivity to be central to her work. This involved making explicit the processes by which her data and analyses are produced, acknowledging her personal interests and values as a researcher and the way in which these influenced the research process. Unlike Sloboda and Howe’s approach, which does not mention the role of the interviewer in eliciting particular responses from the interviewees, Borthwick acknowledged that the interview data she collected are negotiated between her and the interviewee at a particular point in time. As part of this, she discussed her own family background and how this may have influenced her in the development of her questionnaires and the way in which she approached her interviewees. Basing her work on individuals’ accounts rather than attempting to produce an objective statement, Borthwick compared the interpretation of her interviewees’ experiences with looking for meaning in a text. She assumed that what was said by the participants was significant, and that there was some relationship between this and their more enduring beliefs or constructs. She drew on a particular analytical technique which has emerged out of theoretical work on social constructionism and reflexivity—in this case, interpretative phenomenological analysis (IPA). Used extensively by Smith (1996, 1999), this technique is based on the assumption that interviewees do not express all their thoughts and feelings, so that what they say must be interpreted in the light of other observations of their behavior. The procedure involves examining transcripts and other forms of data for themes. The researcher does this pragmatically, making summaries of interviews, lists of associations and potential connections between them. Main themes and subthemes are created and then discussed, the aim being to produce a “grounded analysis” — that is, an analysis based in and emerging from the data. Smith’s work has been in the area of health psychology, and the objective entity of the patient’s body provides a solid backdrop onto which subjective psychological accounts of the physical processes can be projected. In the same way, Borthwick used the musical instrument and progress on it (assessed through examination results and reports from teachers) as the evidence in terms of which she could interpret the interviewees’ accounts of themselves, their relationship with the instrument, and the role of others in shaping their musical development. The advantages of IPA are that it produces very detailed data specific to a particular group or individual, and that it facilitates an intimacy with the data (and arguably the participants), in a way that formally structured quantitative analyses do not; at the same time, the technique of triangulation serves to counteract what might otherwise be seen as the excessive informality or subjectivity of the approach. While the analysis of verbal or written material may appear to be less technical than quantitative analysis, and in that sense easier, qualitative techniques require sensitivity and critical insight that depend on a rigorous training: qualitative research is by no means an easy option, as it is heavily dependent on the researcher’s skills with text and interpretation, and the ability to write about his or her experiences. This is not to say that quantitative research does not also require skills of interpretation, but the results obtained from quantitative research are usually contained within a more generally agreed interpretative framework, and make reference to more formally defined notions of significance.

65

66

EMPIRICAL MUSICOLOGY

These considerations demonstrate the importance of finding an appropriate theoretical and practical approach to the issues and materials of a research topic. Nowhere is this more evident than in the subject of the next section: research into the influence of social factors on the shaping of musical tastes.

Social Factors in Musical Taste People nowadays have access to a huge range of musics from across the world, through both live performance and broadcast or recorded media. Yet, despite this global potential, research demonstrates what most of us probably suspect already — that different types of music appeal to people of differing social groups. As Russell (1997) points out, the audience at a Country and Western concert will probably have little in common with the members of an opera house audience. People’s tastes are heavily influenced by a huge variety of social factors, and a primary aim of research in this area has been to pinpoint the principal sources of influence. The results have typically shown consistent social class, age, and gender biases (DiMaggio and Useem 1978, Pegg 1984, Abeles and Chung 1996, Russell 1997). Research into musical tastes has generally made use of surveys or questionnaires, in the same way as the large-scale study by Sloboda and his colleagues. The techniques Sloboda et al. employed were straightforward: structured questionnaires consisting of closed questions designed for the specific groups targeted, with the completed questionnaires analyzed simply by counting the frequency of different responses. The researchers did not attempt to survey large numbers of the population, but they drew upon representative samples from different age groups and both sexes, in order to give focus to their research while at the same time avoiding undue narrowness. But social psychological research in the domain of musical taste has not always involved simply asking people to rate or report how much they like or dislike a particular kind of music. Surreptitious methods of data collection have also been used, and these raise a further set of issues. A concrete example is provided by North and Hargreaves’s (1996) study of responses to music in a university canteen.

Surreptitious Methods of Data Collection The aim of North and Hargreaves’s study was to test a theory of aesthetic response to music proposed by Berlyne (1971). The theory proposes a relationship between structural complexity and liking (listeners prefer music that is neither too simple nor too complex), so North and Hargreaves used music with low, moderate, and high levels of complexity. Since the study drew on university students as participants, they decided to play a kind of music that they believed that the students might recognize: New Age. They undertook a series of pilot investigations in order to ensure that the specific musical materials were unfamiliar (so that previous knowledge of a particular track could not influence judgment), but nonetheless readily recognizable as belonging to the same genre. This meant that comparisons between the different musical examples would be valid, ideally reflecting only differences in complexity. In order to ensure appropriate differences in musical complexity, the pilot investiga-

Music as Social Behavior

tions were also designed to confirm that the materials conformed with Berlyne’s general definition of aesthetic complexity: high-complexity excerpts were “unpredictable, erratic, and varied,” while low-complexity excerpts were “predictable, simple, and uniform.” All excerpts for the pilot study were 30 seconds long, and were selected to be representative of the piece from which they were taken. In the experiment proper, they played the tracks over loudspeakers in the canteen, and in a random order. The authors were concerned that the study should be as “ecologically valid” (i.e., naturalistic) as possible. For this reason the questions about musical preference were concealed within a general questionnaire survey that asked the students about the canteen environment. So as to give credibility to this apparent aim, the experimenters went around the canteen, asking the students what aspects of the environment they might like to change. The surreptitious assessment of the students’ musical preferences was accomplished by noting how readily they cited the music as an aspect of the environment which they might like to change, or how much they liked it; the researchers also recorded the times at which each questionnaire was filled in, so that they could cross-refer it to the music being played at that time. The results were in line with Berlyne’s theory, with moderately complex music being preferred. The surreptitious nature of the research technique enabled the researchers to show that as dislike of the music increased, the music became more salient. Had the experimenters asked the students about the music in a more direct fashion, they would have drawn attention to the music, eliciting a more self-conscious assessment of these students’ musical preferences. After this brief consideration of work on audience behavior, the next section considers research into social interaction between performers, with a specific focus on how people work together to create a musical performance.

Ensembles as Musical and Social Groups There have been only a few studies exploring group processes in music, most of them either small-scale qualitative investigations or single case studies. The principal reason for this is that in order to assess how a group operates, it is necessary to have a detailed account of their daily practices and operating procedures, and qualitative approaches permit this depth of enquiry. Murnighan and Conlon’s (1991) work is the closest to a more generalizable investigation, combining quantitative analysis with qualitative techniques. They initially contacted 21 string quartets by letter and phone, of which all but one agreed to take part. The study took the form of semistructured interviews with individual quartet members, which lasted between 45 minutes and four hours. The data were then quantified where possible, and quotations were used as a basis for discussions between the experimenter and the players, in which thematic issues were developed. From the broad range of data collected, it seems that the success of a quartet depends on social factors within the group (such as the way in which the second violinist interacts with the other players), as well as on issues of skill and repertoire. Murnighan and Conlon’s work emerges out of the same tradition exemplified

67

68

EMPIRICAL MUSICOLOGY

by Sloboda and his colleagues’ work on musicians’ biographies. An alternative to the questionnaire approach is to study ensemble behavior in a more direct manner, an example of which is a study by Davidson and Good (2002); this used various kinds of directly recorded data to focus on the rehearsing and performing of a student string quartet.

A Direct Approach to Ensemble Research Davidson and Good (2002) studied the musical and social processes of this quartet both in their general interactions with one another, and when rehearsing and performing. They did this by adopting a number of research techniques. First, in individual semistructured interviews, they asked the members of the quartet to give background details about how the ensemble had formed. Then questions were asked about who led both the musical and the social interactions between the quartet members (how the music should be played, how repertoire was chosen, which individuals were dominant, and so on); this revealed the players’ own views about the dynamics of the group. Then, in a second stage, the researchers simply placed a video camera in the rehearsal room. They analyzed the recordings using the following categories, for which detailed criteria were drawn up: • Social conversation (general topics related to friendship, jokes, etc.). • Nonverbal social interaction (related to nonmusical issues, and including physical contact, gestures, degree of proximity, looking behaviors, etc.). • Musical conversations (discussions about technical or expressive points in the music). • Nonverbal musical interactions (gestures demonstrating a musical purpose: coordinating entrances and exits, expressive gestures for particular passages, etc.). • Musical interactions (dynamics, timing profiles, and when the music starts and stops). The two authors carried out the analysis together, checking with one another to confirm or disconfirm their ideas. Examples of each of the categories were also checked by an independent evaluator as a measure of the reliability of the researchers’ assessments. Performance data—a video recording of the concert for which the quartet had been rehearsing—were analyzed in the same manner as the rehearsal data. All members of the quartet were then asked to view the rehearsal and performance videos, giving feedback on the nature of their musical and social behavior. These discussions were also videotaped, so that they could be aligned with the original video footage, and they were themselves subjected to a similar form of analysis. Finally, when all this material had been analyzed, the researchers took their analysis back to the four players for further commentary. This was in part to ensure that the analysis was a fair representation of the players’ experiences, but more importantly, it served as an opportunity for further issues to emerge. One of these issues (the dominance of the male second violinist and a sexual dynamic between him and the female first violinist) caused a little embarrassment, but all players spoke about the role this individual had in the group, and how it had shaped both the general

Music as Social Behavior

conversation and the first violinist’s style of leadership. Had the researchers not fed their analysis back to the players, this issue might never have been clarified. The high level of participation in this project by the players made it a dynamic enterprise. The data have a degree of richness and detail that is often missing in social studies. The advantages of this kind of research are that the data are collected under real-world conditions; that the participants are able to give their interpretations and provide points of clarification for the analysis; and that it provides results that are very specific to the individuals concerned, through accessing their individual and collective experiences. The disadvantages are the cumbersome and timeconsuming data collection and feedback procedures. The research described above has hinted at the role of the individual in the formation of group dynamics and so in the creation of a musical performance. It is appropriate, therefore, to conclude this chapter by considering how the differences between individual musicians have been researched, and what kinds of traits have been found.

Individual Differences In general psychological research, there was a surge of interest during the 1930s and 40s in the extent to which individuals differed from one another, and how the differences between them could be categorized (see Atkinson, Atkinson, Smith, Bem, and Hilgard 1990 for details). This research was fueled by the need to find measures of skill and ability for recruitment to the armed services. Broadly speaking three kinds of personality test were developed: questionnaires or inventories, projective tests, and objective tests. These can be classified under two headings (Kline 1994): nomothetic, that is looking for aspects of personality that are common to different individuals; and idiographic, attempting to assess those that are peculiar to the individual. Personality inventories tend to be nomothetic, while projective tests are idiographic; objective tests may be of either type. The examples below will give a flavor of these tests.

Personality Questionnaires or Inventories Personality questionnaires consist of sets of items (statements or questions) relevant to the personality variables that the test measures, and to which subjects have to respond. When designing a test, the choice of variables to be examined is a complex matter, and researchers have shown that there is a vast number of personality traits which might be measured. Some have argued that there could be as many personality traits as there are descriptive terms for behavior. However Digman (1990) has shown that there is a considerable degree of overlap between the variables, as a result of which five broad traits, together with about 50 independent traits, account for a large proportion of the variation between individuals. The five broad traits have been studied for many decades and inventories have been designed to access them; these usually comprise items which have to be rated as true or false about preferred forms of behavior—for instance, “I always lock my

69

70

EMPIRICAL MUSICOLOGY

door at night.” The traits are labeled as extraversion, agreeableness, conscientiousness, neuroticism, and openness (see Goldberg 1993 for more details). As for the independent traits, a number of tests have again been developed, but the most relevant here is Cattell’s Sixteen Personality Factor questionnaire (Cattell, Eber and Tatsuoka 1970): this is now widely used and standardized, and it is Cattell’s inventory that has been used in the most well-known research on musicians’ personalities. This research is summarized in Kemp (1996), and shows that musicians have different types of personalities from nonmusicians. Within musicians, composers are found to be more imaginative and sensitive than others, while brass players are more outgoing and boisterous than string players. As a group, however, there is a general tendency for musicians to be introverted, and they display much higher levels of androgyny in their personalities than nonmusicians. Kemp uses these findings as the focus for a discussion of whether it is an individual’s personality per se, or the social factors surrounding him or her, that leads to these rather uniform behaviors. This is, of course, the central issue: what can these results usefully demonstrate? Clearly they can show that there are group characteristics (the brashness of brass players, for instance), and these can be compared with those of other groups (visual artists, say, or white middle class males). The data do not, however, enable researchers to say why such behaviors occur: the question which Kemp addresses has the look of chickenand-egg about it. The details of such personality measures need to be carefully studied before empirical work is undertaken, but many investigators find these standardized tools very useful and they have a long track record. They are also publicly available and easy to score, and have been extensively tested for reliability. Moreover, because they are standardized, an individual’s personality can easily be tested by administering the same inventory on a number of occasions, and the resulting profile can be compared to norms derived from the typical profiles of various sample groups. Kemp’s book is a good starting point for exploring such inventories; the detailed discussion of his own research shows how the standard questionnaires can be modified for use with musicians, and how specific data can be tested against the norms. There are, however, two main disadvantages of the inventory technique. First, items have to be short in order to cover all the dimensions of the personality, and such brevity can lead to simplistic statements which fail to capture the complexities of an individual and his or her behaviors. Second, responses can be faked. For instance, if you complete an inventory in a job interview which bluntly asks whether or not you are shy, unreliable or erratic, you may not be inclined to give an honest answer.

Open-Ended or Projective Tests Projective tests emerged from Freudian psychodynamic theory, with its emphasis on imaginative production and its attempt to gain insights into subconscious or hidden aspects of the personality. A typical test of this kind is Rorschach’s inkblot (Rorschach 1921; see Kline 1994, for details). In the test a series of 10 cards, each with a different size and color of inkblot, is shown to the participant and she or he is asked to report everything that each inkblot resembles. The responses are examined according to three main categories:

Music as Social Behavior

• Location (whether all or part of the inkblot is referenced) • Determinants (whether it is shape, texture, or shading that is reported as being important) • Content (what the blot represents) Although some systematic criteria have been applied to the response categories in order to allow for comparisons between individuals, the responses typically provide a qualitative impression of the participant’s general disposition, for instance in terms of competition or cooperation, or a focus on positive or negative images. The advantages of projective tests are that they are a rich source of data. They seem particularly useful in the exploration of subtle aspects of personality that cannot easily be categorized and described, such as an individual’s contrariness. However, these are highly subjective interpretations, and although the rationale is that the participant’s inner world is tapped, guiding a person through a test, and the subsequent interpretation of the data, can be problematic unless the questioner has psychoanalytical skills.

Objective Tests The essential feature of objective tests is that their purpose cannot be guessed by subjects. Typically questionnaires are used in these tests, but the data are not the answers to the questions: rather it is the behavior accompanying the completion of the questionnaire that counts. For example, the degree of acquiescence demonstrated in answering oral questions might be used as a measure of the participant’s anxiety level. Another measure of anxiety level is the “Fidgetometer,” which uses a special chair with electrical contacts which are sensitive to the subject’s movements; a score is derived by calculating the number of movements over a fixed amount of time. Objective tests seem appealing because they are hard to “fake,” but as Kline (1994) points out, despite his own efforts and those of other psychologists interested in individual testing, the validity of these objective tests has yet to be ascertained. In summary, individual differences can be tested in a number of ways. The essential problem facing the researcher is finding a tool which will access an individual’s traits in a manner that is both sensitive and reliable. From the discussion above, it is clear that there are problems as well as advantages with all the different research techniques available.

Personality and Preferences In addition to looking at personality per se, tests can be used to examine the way in which individuals’ preferences and tastes are affected by their personality. An example of this in music is Rawlings and Ciancarelli’s (1997) use of the Neuroticism, Extraversion, Openness (NEO) Personality Inventory (developed by Costa and McCrae 1985), to explore which types of personality correlate with a preference for specific types of music. Given the accessible nature of this study, it is a useful model of how individual difference research can be carried out in musical contexts. Rawlings and Ciancarelli employed 10 categories of question, each made up of several items that identify specific types of music. In framing their questions they made

71

72

EMPIRICAL MUSICOLOGY

use of a standardized measure of musical preference (Litle and Zuckerman’s Music Preference Scale, 1986), to which the researchers added more recent types of music such as “acid jazz” to update the measure, basing their modifications on discussions with representatives of a large music store. Working from a typed questionnaire sheet, participants rated how much they liked each item. The resultant data were analyzed using the NEO Personality Inventory; typical findings were that females liked popular music styles more than males did, and that individuals who scored strongly on the personality dimension “openness” enjoyed a wide range of musical styles.

Conclusion This chapter has sampled just some of the wide variety of techniques available to address research into music as social behavior. But a few key issues emerge which apply to all such work, and these might be summarized as follows. In the first place, the quantity of research in music behavior is quite limited, which means that researchers need to be creative and open-minded in their approach. In other words, do not be put off by a lack of precedent: some of the most engaging and penetrating research emerges out of lateral thinking and an eclectic approach to empirical design. At the same time, you need to be systematic, persistent, and reflexive: it is essential to apply the appropriate research technique correctly, in the sense of fully understanding what data are being collected, and how these data may be analyzed. The research technique you adopt will determine how explicitly you set out your hypotheses, or how far you follow up emergent themes; however it is always important to look for ways of examining the data as fully as possible, and to be aware of possible problems. Constant reflection on working processes is necessary if the result is to be coherent. Finally, analyzing music as social behavior means analyzing how people engage with music and with one another at individual, group, and societal levels, and that in turn means that ethical considerations cannot be ignored. Most academic institutions, health authorities, and other public sector institutions in which research takes place run ethics committees; these monitor proposed research projects so as to ensure that no one is exploited, mistreated, or deliberately subjected to physical or psychological discomfort. If you are going to undertake social research, make sure your proposal is seen and approved by an ethics committee, or failing that by at least three experienced researchers. Getting the design right in the planning stages can save a lot of anxiety later. Exploring music and social behavior is both stimulating and challenging. Face that challenge with energy, an open mind, and sensitivity to others, and your research could make an important contribution to an ever-expanding field. Notes 1. For a discussion of independent and dependent variables, see chapter 9, this volume. 2. A comparison may be drawn with the participant-observation approach common in ethnomusicology; see chapter 2, this volume.

Music as Social Behavior

References Abeles, H. F., and Chung J. W. (1996). “Responses to music,” in D. A. Hodges (ed.), Handbook of Music Psychology, 2nd ed. San Antonio: IMR Press, 285–342. Atkinson, R. L., Atkinson, R. C., Smith, E. E., Bem, D. J., and Hilgard, E. R. (1990). Introduction to Psychology. Fort Worth: Harcourt Brace Jovanovich College Publishers. Berlyne, D. E. (1971). Aesthetics and Psychobiology. New York: Appleton Century Crofts. Borthwick, S. J. (2000). “Parenting Scripts: The Pattern for a Child’s Musical Development.” Ph.D. dissertation, University of Sheffield. Borthwick, S. J., and Davidson, J. W. (2002). “Developing a child’s identity as a musician: A family ‘script’ perspective,” in R. MacDonald, D. J. Hargreaves and D. Miell (eds.), Musical Identities. Oxford: Oxford University Press, 60 – 78. Byng-Hall, J. (1995). Rewriting Family Scripts. London: Guilford Press. Burland, K., and Davidson, J. W. (2001). “Investigating approaches to musical composition with eleven year olds.” Research Studies in Music Education 16: 46–56. Cattell, R. B., Eber, H. W., and Tatsuoka, M. M. (1970). Handbook of the Sixteen Personality Factor Questionnaire. Champaign, Ill.: Institute for Personality and Ability Testing. Costa, P. T., and McCrae, R. R. (1985). The NEO Personality Inventory Manual. Odessa, Fla.: Psychological Assessment Resources. Davidson, J. W., and Borthwick, S. J. (2002). “Family dynamics and family scripts: A case study of musical development.” Psychology of Music 30: 121–136. Davidson, J. W., and Good, J. M. M. (2002). “Social and musical co-ordination between members of a string quartet: An exploratory study”. Psychology of Music 30: 186–201. Davidson, J. W., Howe, M. J. A. and Sloboda, J. A. (1995 / 6). “The role of parents and teachers in the success and failure of instrumental learners”, Bulletin for the Council of Research in Music Education 127: 40 – 45. Davidson, J. W., Howe, M. J. A., Moore, D. M., and Sloboda, J. A. (1996). “The role of parental influences in the development of musical ability,” British Journal of Developmental Psychology 14: 399 – 412. Davidson, J. W., Howe, M. J. A., Moore, D. M., and Sloboda, J. A. (1998). “The role of teachers in the development of musical ability.” Journal of Research in Music Education 46: 141 – 160. Davidson, J. W., and Scutt, S. (1999). “Instrumental learning with exams in mind: A case study investigating teacher, student and parent interactions before, during and after music examination.” British Journal of Music Education 16: 79–95. Davidson, J. W., and Smith, J. A. (1997). “A case study of newer practices in music education at conservatoire level.” British Journal of Music Education 14: 251–269. Digman, J. N. (1990). “Personality structure: Emergence of the five factor model.” Annual Review of Psychology 41: 417 – 440. DiMaggio, P. and Useem, M. (1978). “Social class and arts consumption: The origins and consequences of class differences in exposure to the arts in America.” Theory and Society 5: 141 – 161. Farnsworth, P. R. (1954). The Social Psychology of Music. Iowa: Iowa State University Press. Finnäs, L. (1987). “Do young people misjudge each other’s musical tastes?” Psychology of Music 15: 152–166. Gergen, K.J. and Davis, K. E. (eds. 1985). The Social Construction of the Person. New York: Springer-Verlag. Goldberg, L. R. (1993). “The structure of phenotypic personality traits.” American Psychologist 48: 24 – 34.

73

74

EMPIRICAL MUSICOLOGY

Hargreaves, D. J., and North, A. C. (1999) “The functions of music in everyday life: Redefining the social in music psychology.” Psychology of Music 27: 71–83. Harré, R. (1979). Social Being. Oxford: Basil Blackwell. Harré, R. (1992a). “The discursive creation of human psychology”. Symbolic Interaction 15: 515– 527. Harré, R. (1992b), “Introduction: The second cognitive revolution.” American Behavioral Scientist 35: 5–7. Howe, M. J. A., and Sloboda, J. A. (1991a). “Young musicians’ accounts of significant influences in their early lives: 1. The family and the musical background.” British Journal of Music Education 8: 39–52. Howe, M. J. A.., and Sloboda, J. A. (1991b) “Young musicians’ accounts of significant influences in their early lives: 2. Teachers, practising and performing.” British Journal of Music Education 8: 53–63. Howe, M. J. A., Davidson, J. W. and Sloboda, J. A. (1998a). “Innate gifts and talents: Reality or myth?” Behavioral and Brain Sciences 21, 399–407. Howe, M. J. A., Davidson, J. W., and Sloboda, J. A. (1998b). “Natural born talents undiscovered.” Behavioral and Brain Sciences 21: 432–442. Kanellopoulos, P.A. (1999). “Children’s conception and practice of musical improvisation.” Psychology of Music 27: 175 – 191. Kemp, A. E. (1996). The Musical Temperament. Oxford: Oxford University Press. Kline, P. (1994). “Personality tests,” in S. E. Hampton and A. M. Coleman (eds.), Individual Differences and Personality. London: Longman, 77–96. Lehmann, A. (1997). “The acquisition of expertise in music: Efficiency of deliberate practice as a moderating variable in accounting for sub-expert performance,” in I. Deliège and J. A. Sloboda (eds.), Perception and Cognition of Music. Hove: Psychology Press, 161 – 187. Litle, P., and Zuckerman, M. (1986). “Sensation seeking and music preferences.” Personality and Individual Differences 20: 33 – 45. Murnighan, J. K., and Conlon, D. E. (1991). “The dynamics of intense work groups: A study of British string quartets,” Administrative Science Quarterly 36, June 1991: 165–186. North A. C., and Hargreaves D. J. (1996). “Responses to music in a dining area.” Journal of Applied Social Psychology 24: 491 – 501. Oppenheim, A. N. (1992). Questionnaire Design, Interviewing and Attitude Measurement. New York: Pinter. Pegg, C. (1984). “Factors affecting the musical choices of audiences in East Suffolk,” Popular Music 4: 51–74.. Rawlings, D., and Ciancarelli, V. (1997). “Music preference and the Five-Factor Model of the NEO Personality Inventory.” Psychology of Music 25: 120–132. Robson, C. (1993). Real World Research. Oxford: Blackwell’s. Rorschach, H. (1921). Psychodiagnostics. Berne: Hans Huber. Russell, P. A. (1997). “Musical tastes and society,” in D. J. Hargreaves and A. C. North (eds.), Social Psychology of Music. Oxford: Oxford University Press, 141–158. Searle, H., ed. (1966). Hector Berlioz: A Selection from his Letters. London: Gollancz. Sheridan, D., Bloome, D., and Street, B., eds. (1998). Writing Ourselves: Literacy Practices and the Mass-Observation Project. London: Hampton Press. Sloboda, J. A. (1998). “Everyday uses of music listening.” Proceedings of the Fifth International Conference on Music Perception and Cognition, Seoul, Korea, August, 1998, 55–60. Sloboda, J. A., and Howe, M. J. A. (1991). “Biographical precursors of musical excellence: An interview study”. Psychology of Music 19: 3–21.

Music as Social Behavior Sloboda, J. A.., Davidson, J. W., and Howe, M. J. A.. (1994a). “Is everyone musical?” The Psychologist 7: 349 – 354. Sloboda, J. A., Davidson, J. W., and Howe, M. J. A. (1994b). “Musicians: Experts not geniuses.” The Psychologist 7: 363 – 365. Smith, J. A. (1994). “Reconstructing selves: An analysis of discrepancies between women’s contemporaneous and retrospective accounts of the transition to motherhood”. British Journal of Psychology 85: 371 – 392. Smith, J. A. (1996). “Beyond the divide between cognition and discourse: Using interpretative phenomenological analysis in health psychology.” Psychology and Health 11: 261–271. Smith, J. A. (1999). “Identity development during the transition to motherhood: An interpretative phenomenological analysis.” Journal of Reproductive and Infant Psychology 17: 281 – 299. Smith, J. A., Flowers, P., and Osborn, M. (1997). “Interpretative phenomenological analysis and health psychology,” in L. Yardley (ed.), Material Discourses of Health and Illness. London: Routledge, 68 – 91. Smith, J. A., Harré, R., and van Langenhove, L., eds. (1995). Rethinking Methods in Psychology. London: Sage. Weissweiler, E., ed. (1984). The Complete Correspondence of Clara and Robert Schumann. New York: Peter Lang.

75

This page intentionally left blank

CHAPTER 5

Empirical Methods in the Study of Performance Eric Clarke

Introduction From the perspective of musicology, the empirical study of performance has coincided with a move away from the primacy of the score and toward increasing interest in music as performance. The rise of performance studies as a research area has brought a focus on different performance traditions, the nature of performance interpretation and its relationship to analysis, and the legacy of historical recordings (now dating back 100 years) and what it can tell us about changes in performance styles. Musicologists have been interested in empirical studies of performance as ways of documenting what goes on in performance, and for their ability to make performance a concrete object of study with the same tangibility that was previously confined to scores and sketches. Although performance occupies a central position in just about every musical culture, systematic studies of performance go back only to the turn of the twentieth century. The reason for this is the problem of transience: only once methods had been developed to record either the sounds of performance, or the actions of instruments, was any kind of detailed study possible—and so the piano roll, record, magnetic tape, and computer have all played their part at different stages in the short history of empirical studies of performance. Gabrielsson (1999) provides a survey of empirical studies starting with Binet and Courtier’s (1895) study of piano playing, and demonstrates the rapid growth in the field—particularly since about 1980, while the 37 separate chapters in Rink (2002) and Parncutt and McPherson (2002) indicate how active the field continues to be. A significant factor in this development has been the involvement of psychologists, to whom music performance has appealed as an area of study on various grounds. It represents an example of a very sophisticated and complex motor skill, on which there is a wider research literature in psychology; it has affinities with language (about which there has been a great deal of psychological research), but represents a distinct form of nonverbal communication; it provides an opportunity to study rhythmic and other temporal skills at various levels of expertise; and it provides a “window” onto hidden cognitive processes in music. For these and other reasons, and because of the converging interests of psychologists and musicologists, performance research is perhaps the most developed

77

78

EMPIRICAL MUSICOLOGY

area of empirical musicology—though the differing motivations for studying performance have given rise to certain problems, as I discuss at the end of this chapter. Because of the specific issues investigated, and for technical reasons, the great majority of empirical work on performance has looked at keyboard performance. The keyboard offers a number of specific advantages for empirical research: • It provides a ready-made context in which to study the coordination and control of concurrent tasks, because the two hands essentially “do the same thing” (albeit in mirror image). • There is a large and varied repertoire of solo music for the instrument, providing the opportunity to study individual performance in an entirely realistic way, as well as a large ensemble repertoire (including piano duets and fourhand material). • The percussive character of the instrument makes it an effective way to study rhythmic skills, and makes accurate timing analysis from sound files possible (due to the sharp onsets of events). • It is easier and less intrusive to take direct mechanical measurements from the keyboard than from almost any other instrument, due to the physical separation of the instrument from the performer. • Since the early 1980s a range of commercially available keyboard instruments has existed that can be directly monitored by computer, allowing performance data to be captured and analyzed easily and with considerable precision. Since the mid-1980s this range of instruments has included real pianos. The principal focus of this chapter is therefore on the study of keyboard performance.

Landmarks To give a quick overview of how the empirical study of performance has developed, the following is a list of some of the publications that have played an important part in defining the field: • In the early 1930s, Seashore and his research associates developed an extensive research program in music performance at the University of Iowa, much of which is brought together and reported in Seashore’s summative book [1967 (1938)]. This represents the earliest extensive and systematic empirical work on performance, and identified many of the issues that have remained the preoccupations of subsequent research. • Povel (1977) describes expressive timing in a number of Bach harpsichord performances, and Bengtsson and Gabrielsson (1977) in performances of Swedish folk tunes. Together, these represent the first significant publications of the “modern” period of performance research. • Shaffer (1981) is the first substantial paper to report results obtained from direct computer monitoring of the piano. The paper concentrates on timing, coordination, expression, and the cognitive representation of complex movements.

Empirical Methods in the Study of Performance

• Sundberg, Fryden, and Askenfelt (1983) is the first published attempt to produce an artificial model of performance expression, using a collection of separate rules that relate to local features of the music. Todd (1985) is a subsequent attempt to achieve the same goal using only a single rule applied recursively. • Repp (1990a) is the first paper to look at a larger body of performance data, using commercial recordings and extracting performance data from the recorded sound. Since that first paper, Repp has gone on to investigate larger collections of performances, in some cases analyzing over 100 recorded performances of the same work. • Davidson (1993) is the first published work analyzing the visual component of expressive performance. • Rink (1995) represents the first large-scale publication bringing together musicologists and psychologists in the study of performance. With this overall pattern of development in mind, the following sections discuss the different methods that have been used and some of the issues that have been tackled.

Using MIDI to Study Performance As already observed, a survey of the existing empirical work on performance demonstrates that by far the largest body of research has examined data derived from the direct measurement of keyboard performances. Until the mid-1980s, this was only possible by building a specialized technical setup (examples of which are Seashore’s piano camera in the 1930s, and Shaffer’s piano /computer interface designed by him in the 1970s), or by using synthesizer keyboards. From the early 1980s, synthesizer keyboards could be connected to tone generators and computers using a specific digital communications protocol called the Musical Instrument Digital Interface (MIDI), making it possible to record and store on a computer all the keyboard events of a performance. But because MIDI was initially implemented on keyboards with poor touch characteristics and that produced unconvincing approximations to piano sound, little in the way of serious research was possible. In the mid 1980s, however, Yamaha produced first a synthesizer keyboard with more realistically weighted keys, and then a MIDI grand piano—a standard acoustic grand piano fitted with a photoelectric cell system which picked up the movements of the piano’s keys, hammers, and pedals, and translated them into MIDI signals. This was essentially a commercial implementation of the type of system devised by Shaffer (see above), but with the advantage that it used a standardized method of data transfer, and could therefore be connected to any MIDI compatible computer. Soon after this, Yamaha came up with their Disklavier system (which is essentially the same as the MIDI piano, but with the added facility that the piano can be used to play back files, as well as to record them), while Bösendorfer developed their own sophisticated (and expensive) equivalent. Although the following discussion is based on MIDI data from keyboards, it is worth noting that other kinds of MIDI controllers exist and can be used for per-

79

80

EMPIRICAL MUSICOLOGY

formance research. There are MIDI drum pads, wind controllers, guitar, and other string controllers, as well as pitch-to-MIDI conversion systems designed to take input from a microphone and to convert the signal into MIDI information. This last category of device has the potential to extend the direct measurement method to the study of any instrument, including the human voice, and to make it possible to take ordinary sound recordings and convert them to MIDI. However there are serious limitations with existing pitch-to-MIDI conversion systems: they are not yet reliable or robust in performing the conversion and thus provide rather inaccurate and unreliable data, they cannot deal with anything other than a single melodic line, and they are sensitive to any sounds that might mask, or interact with, the instrumental signal. Nonetheless, advances in signal processing technology and software are likely to make this kind of conversion a practical reality in the future. General introductions to MIDI can be found in a number of readily-available sources (e.g., Penfold 1990, Manning 1993). What follows is a simple and non-technical description of only those features of MIDI that are relevant to performance research. MIDI is essentially a digital coding of the features of a controller (i.e., a performance instrument of some kind) into a form that can either be used to manipulate a sound source to which the controller is connected, or simply stored in a file on a computer. Since the keyboard is the only controller represented in the existing research literature, the discussion here is confined to the way in which MIDI encodes keyboard events. Six features of keyboard performance are captured by MIDI:1 1. 2. 3. 4.

The identity of any key that is struck The time at which a note starts The time at which a note stops The velocity of the key press or of the piano hammer as it hits the string (which is directly correlated with the loudness of the sound produced) 5. The time at which either the sostenuto or soft pedal is depressed 6. The time at which either the sostenuto or soft pedal is released These data can be captured and stored by most desktop or laptop computers running any one of a large number of commercially available software packages — usually a sequencer of some kind (common ones include Vision, Cubase, and Logic, as well as the more sophisticated graphical programming environment called Max). Most of the data described above provide information that is completely self-explanatory: the resulting file records which notes are played, at what time, with what velocity, at what time they are released, and what pedaling might affect them. Two further pieces of data are also available, but not quite so directly: interonset interval (IOI), and articulation. The interonset interval of a note, usually taken to be a note’s primary rhythmic property, is defined as the time from the start of any note to the start of the next note in the same part. The identification of a “part” in the definition given here can be tricky, but is equivalent to the rhythmic stream to which a note belongs (the relationship between the onset of a note in one part and the onset of an adjacent note in the other part is usually meaningless as a rhythmic property — though it can indicate something interesting about synchronization). In the case of a note-against-note texture (such as a two-part invention), the definition of each part is clear enough. Other

Empirical Methods in the Study of Performance

textures may be rather more difficult to partition into streams, and may necessitate making pragmatic (rather than strongly theoretically founded) decisions. The second piece of “hidden” data is articulation. The articulation of a note on a keyboard instrument is determined by the relationship between the sounding and silent parts of a note’s time span: a staccato note is one in which the sounding portion of a note’s IOI is shorter than the total IOI (i.e., part of the IOI is silent), and a legato note is one in which the sounding portion is either the same, or greater, than the total IOI—in this last case overlapping with the following note in the same part. The information to determine a note’s articulation, therefore, is found in the relationship between the value of the IOI, and the period during which the key is actually depressed (given by the time difference between features 3 and 2 in the list of six above). This value can be further influenced by pedaling: if the sostenuto pedal is depressed prior to, or during, the sounding part of the note, then the critical value which determines the offset (end) of the note is either the release of either the note or the pedal—whichever is the later. Most sequencer programs store MIDI information in a format that is appropriate for studio use but awkward for analytical purposes. In order to overcome this difficulty, and to provide various analytical and transforming possibilities, a software environment called POCO has been developed that can convert standard sequencer files into a variety of more readable text formats, provides the means to strip out selectively all kinds of information from the original performance file, and allows the transfer of data directly into a variety of statistics and graphics packages. The software can be used via the Web, and its rationale and structure (at an earlier stage of development) are described in Honing (1990).2 At the most basic level it converts the timing of events from the “bars, beats, and ticks” format that most sequencers use into seconds and milliseconds (ms), and the velocity (dynamic) data from the 0 to 127 values that MIDI employs into a scale from 0 to 1. Other than this simple conversion, MIDI velocity data need no transformation: they give a direct and immediately interpretable picture of the variations in loudness over time.3 The IOI values are rather less transparent, since the duration of any note in a performance is a product of the duration specified by the notation combined with any stretching or contraction of that value introduced by the performer. It is these relatively small departures from what can be termed the “canonical value” of a note (as specified by the notation) that tend to be of interest in empirical studies of performance, since they are taken to indicate expressive effects (see below). For this reason, it is not usually the absolute value of the difference between a note’s actual duration and its canonical value that is of interest, so much as its proportional increase or decrease. Take the case of crotchets alternating with pairs of quavers at an indicated tempo of 60 crotchets per minute. If played exactly as notated, each crotchet should last for exactly 1,000 ms and each quaver for 500 ms. Imagine some data from two keyboard performances of these notes with the IOIs shown in Figure 5.1. Two things are evident from these data: neither performance is metronomically precise (the IOIs vary from their canonical values of 1,000 and 500 ms); and as a whole the first sequence is played slower than the indicated tempo (most of the IOIs exceed their canonical value), while the second sequence is played close to the in-

81

EMPIRICAL MUSICOLOGY

Figure 5.1. Sample performance data (in milliseconds) for two performances of a simple rhythmic pattern.

dicated tempo (there is something like a balance of IOIs that exceed and fall short of their canonical values). It is not easy, however, to compare these two sets of data directly: Figure 5.2 shows the raw data in a graphical form, and although it is possible to see that the data of performance 1 tend to lie above those of performance 2 (i.e., the tempo of performance 1 is somewhat slower), the zigzag profile makes it hard to see what is going on. There are various ways to transform the data to bring out specific features and make comparisons between data easier: a common approach is to divide the duration of each note by a number representing the note’s written value, usually expressed as a multiple of some underlying metrical value. For the sequence of crotchets and quavers being considered here, crotchets are represented by the value 2 and quavers by the value 1, and the result of dividing the two sets of data by the appropriate sequence of ones and twos is shown in Figure 5.3. It is now a great deal more obvious that performance 1 is slower than performance 2, and furthermore it is also clear that performance 1 gets slower as it proceeds (it rises in the graph) while performance 2 stays more or less stable in terms of overall tempo. Figure 5.3 is essentially a tempo map, showing the momentary tempo of each note for the two performances—except that what is shown is actually the in-

(ms)

82

Figure 5.2. Raw timing data for two performances of a simple rhythmic pattern.

(ms)

Empirical Methods in the Study of Performance

Figure 5.3. Normalized timing data for performances of a simple rhythmic pattern.

verse of tempo (sometimes called “normalized 101”). Because tempo is a more familiar concept, it is common for a further transformation (inversion of the data)4 to be applied, with the result shown in Figure 5.4. At least one more feature can be extracted from even such simple and smallscale data—a difference in the treatment of the tempo of quavers in the two performances. In performance 1 the second of each pair of quavers (the quavers are data

Figure 5.4. Timing data represented in terms of momentary tempo (in quaver beats per minute).

83

84

EMPIRICAL MUSICOLOGY

points 2 and 3, 5 and 6, 8 and 9 and 11 and 12) is played consistently slower than the first, while in performance 2 it is the other way around, a result that could be used as evidence either for a difference in performance style between the two sets of data, or for a distinction between cognitive strategies in the two performers who were the source of the data.

Expression in Performance: Its Definition and Analysis A central preoccupation in research on performance has been the nature and function of expression. A crucial question is how expression should be defined and characterized, and whether (and how) it might be measured. From some of the earliest work on performance [Seashore 1967 (1938)] has come the idea that expression should be defined as departures from some neutral norm, this norm itself being inexpressive as far as performance is concerned. For Seashore, and many others who have followed a Western, score-based tradition, this has translated into defining expression as departures from the values (principally rhythmic and dynamic) specified in the score. The rationale for this approach is that what makes a performance expressive is what the performer brings to the piece beyond what the composer has specified in the score. Thus expression consists of systematic departures from the explicit notation of the score—whether these departures are deliberate and conscious on the part of the performer or not (for further discussion, see Clarke 1995). A problem with this approach is that it regards the score as “the piece” in a kind of disembodied, ahistorical fashion, apparently divorced from any of the cultural assumptions about how the notation might be understood and interpreted. For example, it was an assumed convention of the French Baroque to play passages of equally notated quavers in an alternating long-short pattern (notes inégales), and to play dotted notes with variable amounts of over- and underdotting. An analysis which concluded that the performer consistently applied an expressive emphasis to the first of a pair of quavers, or to the dotted note in dotted rhythms, might be thought to have missed the point since these “departures” from the notation are implicit within the notation itself, and thus do not constitute the performer’s contribution. However, exactly the same might be said of the overwhelming tendency for performers to slow up toward the end of phrases or sections, an equally pervasive cultural expectation within the Classical and Romantic repertoire—but this phenomenon has become probably the most intensively studied expressive property of performance. The problem comes down to whether expression should be regarded as the specific contribution of an individual performer, or in more social terms—as the combination of widely shared cultural norms with the particular input of the individual performer. If it is the former, then where one places the boundary between cultural norms and individual expression becomes a critical consideration; if the latter, then a more relaxed attitude to the source of the phenomenon (cultural convention or individual intention) is possible. In practice, few researchers have devoted much attention to these issues. A more concrete methodological issue is how to distinguish between random or accidental variations in timing, dynamics, or articulation, and properties that might legitimately be considered expressive. The problem of distinguishing deliberate (hence

Empirical Methods in the Study of Performance

expressive) features from mistakes is problematic: the most widespread solution adopted in the literature is to make use of the principle of reproducibility, and to pay attention only to those features that emerge from the average of a number of repeated performances, rather than individual data points. While this approach is consistent with the standard approach to empirical data used in the behavioral sciences, it runs counter to a fundamental principle in musical performance — the idea that performance is a recreative, rather than reproductive, act: each performance is a specific realization of a piece of music, and there is no reason why any two such realizations should converge toward identity. Variation between notionally “identical” renderings of the same piece cannot therefore be regarded as random, and the average of a set of performances, by collapsing together distinct approaches to the piece, may have little value. An alternative is to use internal consistency within a single performance, and to analyze the data guided by principles of musical coherence to distinguish expressive features from errors. There are still problems here: first, internal consistency within a single performance is subject to the same “uniqueness” argument as applies to separate performances (there may be good reasons to play the same, or a similar, passage occurring in two places with deliberately different expressive features); and second, musical analysis is not an exact science and cannot be relied upon to provide an unequivocal basis for distinguishing between errors and intentions. The approach actually adopted in the published literature has often depended on the nature of the task and the quality of the data. When the performance data come from relatively simple musical materials, collected under controlled conditions with performers who are not of concert standard, repeated performances and groups of subjects have been used, with standard statistical methods (e.g., Clarke 1993). By contrast, when the data are from concert pianists performing concert repertoire, there is often no opportunity to collect repeat performances, and in these cases the authors may depend on the expertise of the performers and the consequent standard of the data to justify the abandonment of standard statistical methods (e.g., Palmer 1996). A further possibility, demonstrated in work by Widmer (2002), is to apply data mining methods to large bodies of data taken from skilled performers in order to extract general principles from unique performances that all belong to a narrowly focused performance style and repertoire.

Example: Chopin’s Prelude, Op. 28 No. 4, in E Minor The data shown in Figure 5.5 come from an analysis of two performances of Chopin’s E Minor Prelude by a concert pianist playing on a MIDI grand piano (Clarke 1995). The performer was a professional pianist who played on a Yamaha MIDI grand piano in one of the teaching rooms of the Music Department at City University, London. The only people present were the pianist himself and three researchers involved in the study. The performer was originally asked to provide three performances of the prelude in succession, but for a variety of reasons (not least his own interest in the study) he ended up playing the piece six times in the space of about an hour. There were no instructions as to how he should play it on any occasion, and between performances he spontaneously provided an idiosyncratic commentary on

85

86

EMPIRICAL MUSICOLOGY

his playing. The performances therefore took place in a rather unusual—even artificial—situation, but one with which the performer seemed entirely comfortable. The data were recorded on a desktop Macintosh computer using a commercial sequencer (Vision), and were processed using POCO and a statistics program (Statview) in the ways that have already been described. The piece consists of a rhythmically differentiated right hand melodic line accompanied for the most part by a constant stream of three-note left hand quaver chords,5 and this immediately raises questions about how the data should be examined and represented. The data from the right hand are an obvious target for analysis, since the right hand (particularly in this kind of texture) is conventionally regarded as the primary carrier of expression, and in this piece consists of a stream of single notes. The disadvantage is that it provides rather sparse and intermittent data (many bars contain only two notes—a dotted minim and a crotchet) compared with the regular quavers of the left hand. The problem with the left hand is that it consists of three- and four-note chords, each note of which has its own dynamic and timing value. One solution is to simplify the left hand data by regarding each chord as a unit, resulting in a single timing and dynamic measure for each chord. There are various ways in which this might be done, but for the purposes of this example the dynamic value for each chord is represented by the average dynamic value of the notes played in the chord, and the IOI for the chord is calculated as the time from the first note of the chord to be struck (whichever note that is) to the first note of the subsequent chord (whether that note is in the same voice or not). The rationale for this is that the most fundamental rhythmic property of any event sequence is its attack-point pattern, and that the primary dynamic attribute is a property of the whole chord regarded as a fused entity. Figure 5.5 thus shows sequences of single data points for the tempo and dynamic level of each chord in the left hand, for each of two performances. The large-scale, two-part shape of the performances can be seen in the tempo and particularly the dynamic data; the bar-by-bar pattern of expressive timing can be plainly seen at the start; and various differences between the expressive shaping of the two performances can also be seen. In the original publication (Clarke 1995), these differences were used to argue for two different interpretations of the piece (articulated in these two performances) relating to an ambivalence in its underlying formal structure. The purpose of the example here has been simply to illustrate the kinds of data that can be gathered and some of the questions that they can be used to address.

The Limitations of MIDI A number of sources of information available from MIDI were not considered in the study described above (pedaling, articulation, synchronization within chords and between parts), while other features of piano performance are impossible to study using MIDI: because MIDI data are derived from the mechanism of the keyboard action and not the sound, anything that relates to acoustical properties of the instrument is excluded from the kind of analysis shown above. This includes a whole variety of

Empirical Methods in the Study of Performance

Figure 5.5. Tempo and dynamic data from two expert performances of the Chopin Prelude in E minor, Op. 28 no. 4.

timbral properties, which are the consequence of complex interactions between simultaneously and consecutively sounding strings and the whole body of the instrument, as well as of pedaling and the specific actions of the hammers and dampers. These are, of course, properties over which a skilled performer exercises considerable control, and may even be regarded as the primary expressive parameters in certain kinds of repertoire: MIDI data from a performance of Steve Reich’s Piano Phase, for example, would entirely miss the resonance and “streaming” effects that take place between the two pianos in a performance of the piece, and are arguably its most important features. Then there are the physical and social dimensions of the performance: neither the interactions with any listeners who may be present, nor a whole variety of visual data (most obviously the movements, demeanor, and facial expressions of the performer) can be considered. Some of these have been studied separately (see below), but there is little work that tries to bring these different sources of information together to provide a more multidimensional view of performance (though see Clarke and Davidson 1998, for a preliminary and partial attempt, and Juslin, Friberg, and Bresin 2001–2002 for a rather different approach). Finally, there is one simple but rather practical limitation to the use of MIDI: the performances have to be new performances given by living performers who are willing to give up their time to give a performance on a MIDI instrument of some sort. This immediately excludes the huge resource and historical perspective that commercial recordings can provide, and also limits research to the study of those performers who are willing to take part in these “research” performances — often re-

87

88

EMPIRICAL MUSICOLOGY

corded in unnatural circumstances (usually without an audience or any sense of occasion) and on instruments that may not be of concert standard. For this reason, methods have been developed for obtaining performance data from existing recordings, as the next section describes.

Performance Data from Recordings The attractions of working with recordings are self-evident: not only is there is no other way to study Cortot or Michelangeli, but you can also deal with instruments other than the piano, and you are working directly on the real artifacts of musical culture. Moreover an enormous heritage of recordings, now going back over a century, is becoming easily accessible to scholars, partly through CD reissues of historical recordings, but also through the efforts of major collections such as the British Library Sound Archive (London). There are, however, significant drawbacks. One is that the data are much more difficult to extract than MIDI data: retrieving useful information is at best laborious, and at worst impossible. The other is that historical recordings present problems of interpretation comparable to those that apply to other forms of documentary evidence, but less well recognized owing to the comparative novelty of this area of research. One such problem is the fact that recording speeds were not standardized in early recordings (and were in any case subject to mechanical variation). Early recording techniques were also highly intrusive: in the mechanical era (up to 1925) players had to crowd around a large horn, and this not only drastically affected the balance between instruments but also disrupted the social environment of normal performance. (Such recordings may be no more true to life than the carefully posed studio portraits of Victorian families, with their suspiciously well-scrubbed and immaculately dressed children.) Ironically, when the technology to make high-fidelity recordings became available, with Decca’s full-frequency-range recordings (ffrr) from 1945 and the adoption of magnetic tape from 1950, it was rapidly used to create recordings of performances that had never been, by means of editing techniques: “it was not uncommon,” writes Day (2000: 26), “for the master-tape for an LP, lasting perhaps fifty or fifty-five minutes, to be the result of 150 splices.” This in no way invalidates the recording as an object for musicological study: it may not be a direct representation of an actual performance, but at the same time such recordings represent one of the principal forms in which music was made available in the twentieth century, and was consumed by an ever-increasing public. The point to be made is simply that there is nothing straightforward or transparent about recordings as historical sources. There have been two approaches to obtaining performance data from existing sound recordings: one has been to try to derive something like the same level of detail that MIDI recordings provide, while the other has been to focus on a slightly coarser-grained approach, usually limited to timing. The first of these two approaches has been used most extensively by Repp in a stream of publications investigating in some cases very large repertoires of recorded performances (e.g., Repp 1998, 1999).

Empirical Methods in the Study of Performance

His method has been to digitize the music and then use a digital waveform editor (SoundEdit16) to display the waveform at a suitable level of temporal resolution. A cursor is positioned at clearly recognizable note onsets and used to label those points, using auditory feedback from playing the waveform to decide on the onsets of any difficult notes. These labels identify the times of note onsets, and are saved to a file for subsequent analysis. The method is primarily designed to retrieve IOIs, but has also been used to measure asynchronies among nominally simultaneous note onsets. A meaningful analysis of dynamic levels is even more difficult to derive from sound recordings than is timing, because of the problem of estimating the effective dynamic level of simultaneously occurring events (multinote chords). Repp has, however, undertaken such a study (Repp 1999) using software (Signalyze) that provides the amplitude envelope of the signal. His approach is to take the peak amplitude of the envelope after the onset of a note (or chord) as the index of its dynamic level, and to make no attempt to separate out different streams (voices) within the texture;6 clearly the appropriateness of this approach depends on the texture of the music, and it would be of dubious value in a polyphonic context. Repp’s method can in principle be used with any instrument or combination of instruments, although it becomes increasingly difficult to use as the texture becomes more complex, or with instruments that have indistinct note onsets (e.g., the voice). It should also be noted that it is laborious and time-consuming, involving a number of operations to extract the data for each single event. A second approach has been to focus on timing at a variety of levels, and to extract the timing information by tapping on some device (usually a computer keyboard, or a MIDI instrument connected to a computer) in synchrony with the music (e.g., Cook 1995, Bowen 1996, and for a study that also considers dynamic shaping see Martin 2002). This “tap along” method has the merits of simplicity and economy, but clearly depends on the precision with which the researcher can synchronize with the music. For relatively large musical units (sections, phrases, sub-phrases, and even bars) this is not a problem, since the synchronization error will be a small proportion of the measured unit. However, for smaller units (beats or single notes) it can become a serious problem, and the method has generally not been attempted below the level of the beat. One reassuring point is that the synchronization error is not cumulative, since it is monitored and adjusted at every unit onset, and a check on the reliability of the method can be made either by making multiple passes at the material, or by enlisting independent judges.7

Example: Beethoven’s Ninth Symphony In a chapter broadly concerned with the relationship between analysis and performance, Cook (1995) explores Furtwängler’s conducted performances of Beethoven’s Ninth Symphony in the context of Schenker’s analysis of the music. The study focuses on tempo in performance, and uses the “tap along” method to extract bar-bybar tempo information for the first movement of two live performances available on CD (the use of live recordings helped to minimize the problems resulting from editing, which are particularly acute in studies of performance timing). The technique

89

90

EMPIRICAL MUSICOLOGY

involves playing the CD in the CD-ROM drive of a computer, and tapping the space bar of the computer keyboard in synchrony with the onset of each bar. The computer then stores the times for the inter-tap intervals in a file, which can be shown as a barby-bar tempo chart. Cook (1995: 114) observes: Because it is sometimes difficult to decide exactly where the downbeat falls, and because of factors of motor control, this process is not entirely reliable. In repeated tests based on the same passage of music, my responses varied by an average of 60 milliseconds, which is around 3 percent of the total duration of each bar; deviations of 100 milliseconds (5 percent) were quite frequent. For this reason it would be foolish to make too much of small transitions which appear on the bar-to-bar level; the data are not sufficiently accurate. But the discrepancies are not cumulative, and this means that inferences regarding the broad shaping of tempo (which is what [the] chapter is about) are robust. The study goes on to use the data to refute the assertion that Furtwängler’s tempo fluctuations were arbitrary and uncontrolled, and to show that his performances correspond to Schenker’s analytical view of the piece. Cook (1995: 109) points out that, simply from listening to the performances, it appears that Furtwängler shapes his phrases, balances his instrumentation, or articulates formal junctures in ways which do match what Schenker says, or that at least seem to belong within the same language of performance that Schenker is talking. But judgments of this sort are inevitably vague and impressionistic, especially in view of the limited sound quality and control over balance that was possible in live recordings around 1950. And close listening, by itself, is even more inadequate when it comes to large-scale tempo modifications; waves of accelerando and decelerando are clearly audible, but it can be difficult to disentangle them from dynamics, articulation, tonal quality, and all the other dimensions that contribute to the energy or tension level of the music. Unlike these other dimensions, however, tempo relationships can easily be measured in an empirical and reasonably accurate manner. As with the previous example, the intention here is not to raise the issues that are the concern of the original study, but simply to point out that the method is viable and useful, and permits a kind of discussion that would be hard or impossible to sustain without the empirical data that the method provides. One of Cook’s main arguments in the original study is that Furtwängler makes use of relatively large-scale arch-shaped tempo profiles that allow him to use tempo for both structural and rhetorical purposes in performance. Without the tempo graphs which the “tap along” method provides, it would be impossible—or at the very least verbally cumbersome and far more open to dispute—to demonstrate the particular characteristics of performance that are used to argue both for the particular audible characteristics of Furtwängler’s performances, and for the existence of a specific history and ideology of conducting that Cook traces back to Wagner and Bayreuth.

Empirical Methods in the Study of Performance

Evaluative and Qualitative Methods Measuring the timing, dynamic, and even timbral properties of performances has produced a wealth of previously unknown information but, as has already been pointed out, it gives only a very partial view of what happens in performance. In particular, such an approach entirely misses the social dimension of performance (the interactions between performers, and between performers and others, discussed in Davidson, chapter 4, this volume). While it is possible to quantify some of these characteristics, there is much to be gained from adopting a qualitative approach.8 A method which owes a lot to work in psychotherapy is to get performers to speak about their own performances, and then to analyze both what they say and what they do. As a research method this “talking analysis” is aimed at discovering the intentions, motivations, and evaluations of one or more performers in relation to their own (or another’s) performance. The data for this kind of approach are usually of two kinds: first, a sound recording of one or more performances, and second, a sound recording of the commentary by one or more of the original performers, or another commentator. The commentary is often made by a person who listens to the original sound recording of the performance, stopping the recording as often as he or she likes to make whatever comments are appropriate, and possibly doing so on more than one occasion. In this way a detailed account can be built up outside the pressure of “real-time” performance. The subsequent treatment of the resulting commentary can take many forms (see Davidson, chapter 4, this volume; Robson 1993), usually concerned in one way or another with a content analysis. An example of this approach is the work of Sansom (1997), who studied musical and personal interactions in free improvisation. The study involved a number of pairs of performers who improvised together, the participants being experienced improvisers who had played music with one another previously. Sansom recorded each improvising duo, and then on subsequent occasions asked each of the two players individually to listen to the recording and to stop it and comment on anything that occurred to them about what was happening or what they had been thinking about at the time. These commentaries were also recorded and then subjected to a content analysis. The following (Sansom 1997, ii: 9) is an extract from the comments provided by one of the members of a guitar and saxophone duo — two players who are very experienced improvisers and have frequently played together in the past: Sort of little funk chords come out there for some reason, it took me quite a lot by surprise. I quite like the sound, I reiterated them a few times . . . I never usually do . . . techniques so recognizable as that, very rarely. . . . Another thing from that, I don’t really play on the beat, I play in the bits in between. . . . And often on playback it sounds a bit weird ‘cos it’s just like a little time gap between you—the same rhythm but not together on it. I was thinking about it there because the melodic thing isn’t important, we were both skidding about anywhere. We both decided to play around with the rhythm and it sounded a bit messy but er . . .

91

92

EMPIRICAL MUSICOLOGY

I really enjoyed that bit . . . sort of juxtaposing really brutish behaviour with really small, er, considerate little things . . . I can’t remember what comes next, I hope it’s something very quiet—we’ll see. . . . Right. Mick didn’t feel like stopping at the same time as I did . . . er . . . obviously a certain energy that I was feeling diminished. I thought it was a good time to stop. At a time when Mick was still on a high, so I started to listen to see what he would do. How far he would go on without me commenting. I’d a fair idea he would go on quite a distance . . . I’d no idea what he’d make it into. . . . The analysis of this and similar passages provided by Sansom’s participants focuses on both musical and interpersonal processes. As the extract demonstrates, the players’ transcripts constitute a complex mix of recollection, current comments, awareness of musical materials, and awareness of interpersonal dynamics. Sansom’s content analysis categorizes the remarks into themes that then form the basis for a discussion of creativity and interaction in ensemble improvisation. The advantage of a qualitative approach such as this is its potential to investigate an enormous diversity of phenomena that the quantitative methods discussed earlier are incapable of addressing, including issues of motivation, intention, reaction, and evaluation. A difficulty, however, is the way in which empirical method is combined with interpretation. Every empirical method has an interpretative component, but in most quantitative methods the premises and boundaries of interpretation are usually fairly well recognized. Qualitative methods, however, have been accepted for a very much shorter time, and have not yet acquired the stability (it may simply be familiarity) of their quantitative counterparts. When a quantitative test gives a certain outcome, and subsequent discussion then interprets that result and places it in the context of other work, it is fairly clear (if one accepts the premises of the method) where the reporting of results stops and interpretation starts. This is much less clearly the case with a qualitative method such as that used by Sansom, where the assignment of verbal data to content categories is itself a strongly interpretative process. The objection leveled at qualitative research of this kind is that it is too speculative — that it sets itself up as empirical, and then goes about its business in a manner that looks more like literary criticism. This objection partly reflects the fact that the interpretative assumptions of most quantitative methods have simply become so deeply embedded as to be invisible, but it remains the case that qualitative methods have yet to attain the systematic and explicit character of empiricism in the eyes of many. Robson (1993) and Smith, Harré, and van Langenhove (1995) provide fuller accounts of this debate.

Analyzing Visual Components of Performance Performance is not only a sonic event, and the visual component of performance offers a rich domain for empirical study. In a number of publications, Davidson (1993, 1994, 1995) showed how observers could make accurate judgments of the expressive properties of performances on the basis of video data. The participants in her

Empirical Methods in the Study of Performance

studies watched and listened to a number of different types of material: normal video with or without sound, video with sound from a different performance overdubbed, and what are called “point-light displays.” These last are video recordings made with the performer wearing small spots or strips of reflective material on the major joints (wrist, elbow, shoulder, hip, knee) and head in strong illumination, and then played back on a monitor with the contrast turned to maximum. The resulting image consists simply of spots of light, with the rest of the body and instrument invisible. Earlier work in psychology (e.g., Runeson and Frykholm 1983) showed that viewers are able to perceive significant attributes (such as gender, physical effort, and intention) from a point-light display, suggesting that this information is conveyed by the dynamics of the moving display. Davidson showed similarly that viewers were able to distinguish accurately between expressionless, normal, and exaggerated performances by a variety of instrumentalists on the basis of as little as two seconds of pointlight video without sound. She also found that, when sound was present, nonmusicians tended to be less influenced by it than by the video image: for example, when a point-light video of an exaggerated performance was overdubbed with the sound of the same person playing in an expressionless manner, nonmusicians tended to rate the performance toward the exaggerated end of the scale — whereas their ratings were toward the expressionless end of the scale when the image was from the expressionless performance and the sound from the exaggerated condition. Other studies by Davidson (e.g., Davidson 1994, 1995) used ordinary video images and were designed to investigate whether a performer made use of a fixed repertoire of expressive gestures in performance, and whether viewers could identify their meaning. The technique involved making video recordings, from a fixed position, of a performer (in this case a pianist) who played a mixture of repertoire. The video recordings were shown to viewers who were asked to identify specific types of movement gesture, and give some indication of what they thought these gestures “meant” in terms of their relationship with the music. A variant of this method was a study (Clarke and Davidson, 1998) that combined the type of MIDI data analysis described earlier in this chapter with a movement analysis taken from a video recording. In this case, the movement analysis involved the detailed tracking of a pianist’s head movements using specialized video analysis equipment. The resulting continuous plot of head position was then compared with MIDI data from the piano in an analysis of the relationship between expressive movement, and expressive timing and dynamics. At a still more detailed level, Sloboda, Clarke, Parncutt, and Raekallio (1998) used video to record the hand movements of pianists as part of a project on keyboard fingering. A camera was mounted directly above the keyboard of a MIDI instrument, with powerful sideways illumination to highlight (with light and shadow) which finger depressed which key. The video recording was then used to match the finger that depressed each key with the corresponding note in the MIDI output. There is also the potential to use rather more sophisticated and automated methods in analyzing movement in performance. A preliminary study by Winold, Thelen, and Ulrich (1994) investigated the bow-arm movements of cellists, using a method which employed an optoelectronic motion analysis system, in which the position in space of various specified points on the player and instrument (just the bow in this case) were

93

94

EMPIRICAL MUSICOLOGY

continuously tracked by computer. The technique has the potential to provide data on the continuous movements of instrumentalists and conductors during performance, and offers the prospect of a much more diverse and detailed study of movement in performance than has so far been undertaken.

Example: The Detection of Expression in Point-Light Displays. Having discovered that viewers can pick up the expressive intentions of performers from point-light displays, Davidson (1994) studied which parts, or combinations of parts, of the body are particularly “informative” in making expressive judgments. The design of this study was that a single performer (a pianist) played a short extract of music in three different performance manners (deadpan, expressive, exaggerated), and with seven different arrangements of point-light patches attached to his body: 1. 2. 3. 4. 5. 6. 7.

All patches: head, right shoulder, right elbow, and both wrists All patches except elbow Head and wrists only Wrists only Head only Elbow only Shoulder only

This is clearly not an exhaustive list of all the possible combinations, but it represents the maximum that could be reasonably investigated given that the pianist had to play the music in each of these point-light arrangements, and in each of the performance manners, resulting in 21 (3 ⫻ 7) performances. Video recordings were made of all 21 performances, and were shown (without sound) in different random orders to a total of 15 observers (graduate music students) who were instructed to rate each performance on a seven-point scale of expressivity (from 1 ⫽ inexpressive to 7 ⫽ highly expressive). The results showed that the observers were reliably able to identify the three different manners of performance; that the different point-light arrangements resulted in significantly different patterns of scores; and that the scores for the three versions were different for each point-light arrangement. From this somewhat complex pattern of findings, one simple discovery is that the head and wrists seem to convey most of the expressive information: the “head and wrists only” arrangement gave results that were essentially identical to those for the “all patches” condition, and even “head only” gave a pattern of scores that was close to that for “all patches”. “Wrists alone,” however, seemed to give very misleading information, with all three performance manners eliciting rather similar (and inappropriately high) expressivity ratings. Some of the observers’ comments indicated that it was the level of activity in the wrists (largely a consequence of the movements necessary simply to play the piece) that led to consistently high ratings. As one subject put it (Davidson 1994: 297–298) “I suspect I’m rating this wrist performance as highly expressive just because there is plenty of action. When wrists are combined with other areas of the body, I look to the whole, and the emphasis on fast motion is reduced for I see an elegant arch of the back and a delicate hand lift and it is from the combined infor-

Empirical Methods in the Study of Performance

mation that I am able to recognize the intention of the piece.” The study, then, shows how relatively simple techniques can be used to investigate how different parts of the body are involved in communicating expression, and what relationship there might be between quantitative and more qualitative factors in this process. Since this empirical work was done, the available technology has developed enormously, and rather than asking the performer to play 21 times, with different point-light arrangements for each, it would now be possible simply to record three performances (deadpan, expressive, and exaggerated), each with the full array of point lights, and then use computer editing facilities to edit out individual point-lights, allowing all possible combinations to be investigated, and ensuring that each performance manner, of which each edit was a variant, really was constant.

Performance Models and Algorithms The availability of powerful desktop computers and associated software has made new empirical methods possible, and has also facilitated the construction of models of musical performance. Artificial models of performance have played a significant part in the development of empirical approaches and have often been closely connected with empirical work. When research in the late 1970s demonstrated that musical performance possessed a variety of systematic features, it became an attractive proposition to see whether these features could be embodied in, and simulated by, an artificial model of some kind: Sundberg, Fryden, and Askenfelt (1983), Clynes (1983), and Todd (1985) represent some of the earlier examples of this kind of work. While the model developed by Sundberg and his coworkers was based on their own intuitions, and makes no claim to be derived from explicit empirical evidence, Todd’s various models (Todd 1985, 1989, 1992) come from analyses of performance data, and have in turn motivated and directed subsequent empirical studies using one or other model as their starting point (e.g., Windsor and Clarke 1997). The function and purpose of such models have been misunderstood by some — at times because of inappropriate claims made by their authors. As with the majority of work in artificial intelligence, the primary purpose of a model is as an explanatory device: the model embodies certain psychological principles, and the aim is to draw conclusions from its successes and failures about the principles on which it is based. The outputs are of comparatively little significance in themselves: a model of expressive performance, for example, has little value as a source of expressive performances, but has the potential to identify important issues in the theory on which it is based. The purpose of this brief discussion is neither to explain in detail how any of the various models work, nor to chart their successes and failures, but simply to note some general characteristics of the approaches that they embody, and their implications for the empirical study of performance. The models in this field can be divided into those that approach the subject by means of a single explanatory principle (e.g., Todd 1985) and those which adopt a “multi-rule” approach (e.g., Sundberg, Friberg, and Fryden 1991; Juslin, Friberg, and Bresin 2001–2002). Todd (1985, 1989, 1992) has proposed a number of variants of a model of expressive timing and dynamics

95

96

EMPIRICAL MUSICOLOGY

based on sensitivity to phrase structure, and implemented by means of an algorithm (a rule system) that uses tempo and dynamic curves based on pendular motion. The algorithm takes the pitches and rhythms of the score as its input, along with a hierarchical phrase structure analysis of the music, supplied by the user.9 A tempo and dynamic curve is applied to each phrase unit at every level of structure, the result being a pattern of tempo and dynamics that directly reflects the hierarchical importance of every phrase boundary; the user can direct the model to work within more or less extreme limits of tempo and dynamic variation, and to pay more or less attention to different levels of the phrase structure. The output of the model is a file specifying the precise temporal and dynamic value (i.e., the “expressive” value) of each note from the original input file. The model developed by Sundberg and his coworkers (see, e.g., Sundberg, Friberg, and Fryden 1991) is rather different in conception and operation. Rather than using a single principle that is responsive to hierarchical structure, it consists of a collection of separate and much more specific rules, some of which operate globally and have no sensitivity to structure, while others are responsive to local features of the music. An example of the former is the rule “the higher, the louder” which increases the dynamic level of a note in proportion to its pitch height; an example of the latter is the “melodic charge emphasis” rule, which increases the duration, dynamic, and degree of vibrato applied to a note according to the harmonic remoteness of the note from its local tonic. Finally, Juslin, Friberg and Bresin (2001–2002) have proposed a model which combines four elements corresponding to the acronym GERM: a set of generative rules (G) that relate musical structure to performance expression; principles of emotional expression (E) in performance; a component that introduces deliberate random variability (R) in timing, intended to simulate the uncontrolled low-level variations that characterize the human motor system; and a component which is intended to capture the motion character (M) of expressive performance. The generative rules in the first element are essentially those developed by Sundberg and his coworkers, but the addition of the three other elements is an attempt to develop a model that is more inclusive than previous attempts, and that accounts for the influence of features other than structure alone in the shaping of performance expression. An initial empirical evaluation of the model, reported in the paper, suggests that the E element is actually the most powerful, but that the idea of four separate but interacting components is indeed a powerful way to model performance expression, leaving the authors to conclude that “by considering all four components of the GERM model together, we will be better able to understand the variability usually found in music performance data” ( Juslin, Friberg and Bresin 2001 – 2002: 109). From an empirical point of view, the value of such models lies in the definite predictions that they make, against which one can investigate the phenomena of real performance. They may represent a rather more interesting baseline for the measurement of expressive deviation than the idealized flat line of the score — a baseline that, after all, has no psychological reality (repeated studies have shown that it is impossible for a human performer to produce a completely deadpan, or expressionless, performance). In a similar manner, Parncutt, Sloboda, Clarke, Raekallio, and Desain (1997) offer an ergonomic model of piano fingering not as a plausible “solution” to

Empirical Methods in the Study of Performance

the problem of how pianists decide on fingerings, but as a baseline of physical constraints against which to assess actual fingering choices - which arguably are influenced by much more than physical convenience. There has been relatively little empirical investigation of the available models of expression, with the possible exception of Clynes’s work which, because of its controversial claims, has attracted rather more attention and investigation than others (e.g., Repp 1989, 1990b, Thompson 1989, Clynes 1990, 1995). Such empirical work as there is has generally aimed to test the models, assessing the success with which they simulate real performance. An alternative approach, however, is use the model as a tool, as a way of highlighting what it is that makes human performances interesting. This is the perspective taken by my final example.

Example: Schubert’s Impromptu in G Flat Major Windsor and Clarke (1997) use Todd’s (1992) model in conjunction with an expert performance of the first 16 bars of Schubert’s G flat major Impromptu in order to investigate different components of expressive timing and dynamics. After recording MIDI data (tempo and dynamics) from an expert performance on a Disklavier piano, Windsor and Clarke attempted to simulate the performance using Todd’s model, with the only inputs being the pitches and rhythms of the score, and a phrase structure analysis. Different simulation attempts varied only in the “weight” given to each level of the phrase structure, on the assumption that if the model has any validity it should be able to approximate the expert performance if an appropriate pattern of weightings could be found. The output from each attempt was compared with the data from the expert performance using correlation analysis, trying with successive simulations to arrive at the best fit (i.e., the highest correlation) between the simulation and the data. The study found that the best fit for the timing data required weighting the lowest levels of the phrase structure (half-bars and bars) more heavily than the middle and high levels; this weighting, however, resulted in a rather poor fit for the dynamic data, which were modeled much better by weighting the middle and high levels more heavily than the lower ones. This has implications both for the principles and design of the model, and for an interpretation of what the performer might be doing. In terms of the model, it demonstrates that the assumption that timing and dynamics are rigidly and directly linked is too simplistic; and in terms of performance expression, it suggests that performers may be using timing and dynamics in different ways to project (or respond to) different aspects of structure. In this performance, timing seemed to operate in response to more local structures, and dynamics to middle-level structures. The simulations, although approximating to the expert performance, captured its features in only a rather partial manner—and the study was equally concerned with those aspects of the performance that the model did not capture (it included a discussion of how these discrepancies might be explained). The advantage of the model, then, is that it highlights those features of the data that are not systematic, or that are systematic according to some different principle. Those same features might have emerged from a comparison with the “flat line” of the score, but they are thrown

97

98

EMPIRICAL MUSICOLOGY

into much sharper relief when the data are analyzed against the background of an attempt to model the generic features of performance by systematic means. The paper concludes (Windsor and Clarke 1997: 149): In this paper, we have demonstrated the way in which a model of performance may act as a tool in the analysis of performance data. Although the model fails to account for every aspect of a human performance, and could possibly be revised in the light of the data collected here, these failures are seen as positive because they highlight different aspects of musical expression. The model provides a baseline that is derived from a strong theoretical position, against which other expressive strategies can be assessed. In its clear and unambiguous modeling of continuous expressive strategies, the model allows one to factor out noncontinuous strategies in a manner not possible when a performance is analyzed in relation to an isochronous score.

Conclusions and Prospects Since the late 1970s, when the “contemporary” period of empirical work on performance began, there have been a number of significant achievements. A much more systematic description of a whole variety of phenomena that were previously only known in broad outline is now available, and as a result there is more extensive recognition of the importance of studying music as performance rather than in the scorebased manner that has up to now been the dominant mode. This interest in the detailed characteristics of performance converges with a dramatically increased interest in the history of recorded performance. The confluence of these two lines of research raises the fascinating prospect of a much more serious and detailed examination of performance style and its historical changes, as well as a way of bridging the divide between the ways in which notated music and unnotated music are studied. If music is studied as performance, then whether it exists in a notated form or not becomes an issue that is separable from how (or whether) the music is studied at all. There have been problems, however. Because of the sheer volume of data that comes from even a relatively short performance, there has been the danger of a lack of perspective. It is easy to become buried in the expressive minutiae of a four-bar phrase and to forget that performance is rather more than tempo curves and dynamic accents. A related problem is that the aims and types of explanation and discussion that are of interest to psychologists working in this field may be of less interest to musicologists, and vice versa. Psychologists are, broadly speaking, more interested in general mechanisms (of motor control, perception, cognition) than in particular instances, while musicology tends to be primarily concerned with investigating phenomena that are more narrowly focused and more specifically located in terms of history, culture, or genre. This has at times meant that the two disciplines, which should have much in common, pass each other like ships in the night. Equally, the repertories that have been investigated have been limited (most of the psychology of music has been concerned with the tonal, metrical concert music characteristic of the period from about 1750 to 1850), and the data have often been

Empirical Methods in the Study of Performance

regarded in historically and culturally impoverished terms — though there has recently been something of a shift here, with Repp’s work on large collections of commercially recorded performances spanning most of the twentieth century (e.g. Repp 1990a, 1992, 1998). Finally, many of the empirical methods described in this chapter have led to a thoroughgoing reification of performance—to a view of performance that treats it as a thing rather than a process. This is an almost inevitable consequence of the way in which temporal phenomena are transformed into static representations of one sort or another (graphs, data lists, written descriptions). Some of the work described in this chapter retains the more dynamic character of performance as a process, but this has been the exception, and the more general tendency is to convert performance, in one way or another, into something that is disturbingly like a score — an irony given that one of the motivations for studying performance in the first place (as observed at the start of this chapter) was to get away from the tyranny of the score. Various prospects for empirical work on performance can be envisaged. More automated methods for deriving data from commercial recordings seem inevitable (arriving on the back of digital signal processing technologies in studio recording and editing, and speech processing research), and these should enable much more extensive analyses of large bodies of recorded performance. This in turn will allow more diverse repertories to be tackled, and different questions to be explored. It will be interesting to see whether the emphasis then shifts from the concern with general mechanisms characteristic of existing published research toward more focused questions relating, for instance, to specific styles, performers, or pieces. Equally, the diversification of methods (the analysis of MIDI data and recorded performances now complemented by video and discourse analyses) promises a more integrated and complementary relationship between quantitative and qualitative approaches, and an overdue recognition of the kind of work that has gone on for many years within ethnomusicology (see Stock, chapter 2, this volume). This, and a greater awareness of social and developmental issues (see Davidson, chapter 4, this volume), signals an opportunity to move beyond the cognitive orientation of the last 20 years into a far more multidimensional approach to the study of musical performance. Notes 1. There are additional features such as pitch bend and other so-called controller values, but these have never been studied in the literature discussed in this chapter. 2. Information about POCO, and instructions for using it via the Web, can be found at http:/ / www.nici.kun.nl / mmm. 3. The issue of how a MIDI value based on the speed of a hammer, or a key depression, relates to the actual loudness of the corresponding note complicates things a little. But most synthesizers and Disklaviers implement an essentially linear relationship between MIDI values and decibels (a measure of loudness that takes into account the nonlinearity of the human auditory system in its response to the physical intensity of a sound source; see chapter 8, this volume). 4. Strictly speaking the transformation expresses each data point as its reciprocal value: value A in milliseconds is transformed into 1,000 / A ⫻ 60, which converts it into tempo expressed in beats per minute (BPM).

99

100

EMPIRICAL MUSICOLOGY

5. The only exceptions are a one-bar break in the left hand at the midpoint of the piece, a similar one-bar break just before the three-chord cadence that concludes the piece, and the cadence itself (which consists of two minims and a semibreve). 6. For a discussion of auditory streaming see chapter 8, this volume. 7. The option of using independent judges is usually unrealistic, since in order to produce a good set of “tap along” data, a person has to get to know the music very well so as to be able to anticipate and track tempo fluctuations. 8. The distinction between quantitative and qualitative approaches can be simply expressed as the difference between phenomena that you can measure or give a value to (duration, dynamic, the number of occurrences of a word, the size of a movement, etc.), and those for which this is impossible (the semantic content of speech, the shape of a movement). Further discussion can be found in chapter 9, this volume. 9. There is no attempt within Todd’s model to simulate, or model artificially, the process by which structure is assigned to a score.

References Bengtsson, I., and Gabrielsson, A. (1977). “Rhythm research in Uppsala,” in Music, Room, Acoustics. Stockholm: Royal Swedish Academy of Music, 17: 19–56. Binet, A., and Courtier, J. (1895). “Recherches graphiques sur la musique.” L’Année Psychologique 2: 201– 222. Bowen, J. A. (1996). “Tempo, duration, and flexibility: Techniques in the analysis of performance.” Journal of Musicological Research 16: 111–156. Clarke, E. F. (1993). “Imitating and evaluating real and transformed musical performances.” Music Perception 10: 317 – 343. Clarke, E. F. (1995). “Expression in performance: Generativity, perception and semiosis,” in J. Rink (ed.), The Practice of Performance. Cambridge: Cambridge University Press, 21–54. Clarke, E. F., and Davidson, J. W. (1998). “The body in performance,” in W. Thomas (ed.), Composition — Performance — Reception. Studies in the Creative Process in Music. Aldershot: Ashgate Press, 74 – 92. Clynes, M. (1983). “Expressive microstructure in music, linked to living qualities,” in J. Sundberg (ed.), Studies of Music Performance, Stockholm: Royal Swedish Academy of Music, 39: 76– 181. Clynes, M. (1990). “Some guidelines for the synthesis and testing of pulse microstructure in relation to musical meaning.” Music Perception 7: 403–422. Clynes, M. (1995). “Microstructural musical linguistics—composers pulses are liked most by the best musicians.” Cognition 55: 269 – 310. Cook, N. (1995). “The conductor and the theorist: Furtwängler, Schenker and the first movement of Beethoven’s Ninth Symphony,” in J. Rink (ed.), The Practice of Performance. Cambridge: Cambridge University Press, 105–125. Davidson, J. W. (1993). “Visual perception of performance manner in the movements of solo musicians.” Psychology of Music 21: 103 – 113. Davidson, J. W. (1994). “What type of information is conveyed in the body movements of solo musician performers?” Journal of Human Movement Studies 26: 279–301. Davidson, J. W. (1995). “What does the visual information contained in music performances offer the observer? Some preliminary thoughts,” in R. Steinberg (ed.), Music and the Mind Machine: The Psychophysiology and Psychopathology of the Sense of Music. Berlin: Springer Verlag, 105 – 113.

Empirical Methods in the Study of Performance Day, T. (2000). A Century of Recorded Music: Listening to Musical History. New Haven: Yale University Press. Gabrielsson, A. (1999). “The performance of music,” in D. Deutsch (ed.), The Psychology of Music, 2nd ed. San Diego, Calif.: Academic Press, 501 –602. Honing, H. (1990). “POCO: An environment for analysing, modifying, and generating expression in music.” Proceedings of the 1990 International Computer Music Conference, 364 – 368. San Francisco: Computer Music Association. Juslin, P. N., Friberg, A., and Bresin, R. (2001 – 2002). “Towards a computational model of expression in music performance: The GERM model.” Musicae Scientiae, special issue on Current Trends in the Study of Music and Emotion: 63–122. Manning, P. (1993). Electronic and Computer Music, 2nd ed. Oxford: Oxford University Press. Martin, S. (2002). “The case of compensating rubato.” Journal of the Royal Musical Association 127: 95 – 129. Palmer, C. (1996). “Anatomy of a performance: Sources of musical expression.” Music Perception 13: 433 – 454. Palmer, C. (1997). “Music performance.” Annual Review of Psychology 48, 115–138. Parncutt, R., and McPherson, G. E. (eds., 2002). The Science and Psychology of Music Performance. Creative Strategies for Teaching and Learning. Oxford: Oxford University Press. Parncutt, R., Sloboda, J. A., Clarke, E. F., Raekallio, M., and Desain, P. “An ergonomic model of keyboard fingering for melodic fragments.” Music Perception 14: 341–382. Penfold, R. A. (1990). Practical MIDI Handbook, 2nd ed. Tonbridge: PC Publishing. Povel, D-J. (1977). “Temporal structure of performed music. Some preliminary observations.” Acta Psychologica 41: 309 – 320. Repp, B. H. (1989): “Expressive microstructure in music: A preliminary perceptual assessment of four composers’ ‘pulses’.” Music Perception 6: 243–274. Repp, B. H. (1990a). “Patterns of expressive timing in performances of a Beethoven minuet by nineteen famous pianists.” Journal of the Acoustical Society of America 88: 622–641. Repp, B. H. (1990b). “Further perceptual evaluations of pulse microstructure in computer performances of classical piano music.” Music Perception 8: 1–33. Repp, B. H. (1992). “Diversity and commonality in music performance: An analysis of timing microstructure in Schumann’s Träumerei.” Journal of the Acoustical Society of America 92: 2546 – 2568. Repp, B. H. (1998). “A microcosm of musical expression: I. Quantitative analysis of pianists’ timing in the initial measures of Chopin’s Etude in E major.” Journal of the Acoustical Society of America 104: 1085 – 1100. Repp, B. H. (1999). “A microcosm of musical expression: II. Quantitative analysis of pianists’ dynamics in the initial measures of Chopin’s Etude in E major.” Journal of the Acoustical Society of America 105: 1972 – 1988. Rink, J. (ed., 1995). The Practice of Performance. Studies in Musical Interpretation. Cambridge: Cambridge University Press. Rink, J. (ed. 2002). Musical Performance. A Guide to Understanding. Cambridge: Cambridge University Press. Robson, C. (1993). Real World Research. A Resource for Social Scientists and PractitionerResearchers. Oxford: Blackwell Publishers. Runeson, S., and Frykholm, G. (1983). “Kinematic specification of dynamics as an informational basis for person-and-action perception: Expectations, gender, recognition, and deceptive intention.” Journal of Experimental Psychology: General 112: 585–615.

101

102

EMPIRICAL MUSICOLOGY

Sansom, M. J. (1997). “Musical Meaning: A Qualitative Investigation of Free Improvisation.” Unpublished Ph.D. thesis, University of Sheffield. Seashore, C. E. [1967 (1938)]. The Psychology of Music. McGraw-Hill. (Republished by Dover Books, New York, 1967.). Shaffer, L. H. (1981). “Performances of Chopin, Bach and Bartók: Studies in motor programming.” Cognitive Psychology 13: 326 – 376. Sloboda, J. A., Clarke, E. F., Parncutt, R., and Raekallio, M. (1998). “Determinants of fingering choice in piano sight-reading.” Journal of Experimental Psychology: Human Perception and Performance 24: 185 – 203. Smith, J. A., Harré, R., and van Langenhove, L. (eds., 1995). Rethinking Methods in Psychology. London: Sage. Sundberg, J., Fryden, L., and Askenfelt, A. (1983). “What tells you the player is musical? An analysis-by-synthesis study of music performance,” in J. Sundberg (ed.), Studies of Music Performance, Stockholm: Royal Swedish Academy of Music 39: 61–75. Sundberg, J., Friberg, A. and Fryden, L. (1991). “Common secrets of musicians and listeners: An analysis-by-synthesis study of musical performance,” in P. Howell, R. West, and I. Cross (eds.), Representing Musical Structure. New York and London: Academic Press, 161 – 197. Thompson, W. F. (1989). “Composer-specific aspects of musical performance: An evaluation of Clynes’s theory of pulse for performances of Mozart and Beethoven.” Music Perception 7: 15–42. Todd, N. P. (1985). “A model of expressive timing in tonal music.” Music Perception 3: 33–58. Todd, N. P. (1989). “A computational model of rubato.” Contemporary Music Review 3: 69–89. Todd, N. P. (1992). “The dynamics of dynamics: A model of musical expression.” Journal of the Acoustical Society of America 91: 3540 – 3550. Widmer, G. (2002). “Machine discoveries: A few simple, robust local expression principles.” Journal of New Music Research 31: 37 – 50. Windsor, W. L., and Clarke, E. F. (1997). “Expressive timing and dynamics in real and artificial musical performances: Using an algorithm as an analytical tool.” Music Perception 15: 127– 152. Winold, H., Thelen, E., and Ulrich, B. D. (1994). “Coordination and control in the bow arm movements of highly skilled cellists.” Ecological Psychology 6: 1–31.

CHAPTER 6

Computational and Comparative Musicology Nicholas Cook

Introduction The middle of the twentieth century saw a strong reaction against the comparative methods that played so large a part in the disciplines of the humanities and social sciences in the first half of the century, and musicology was no exception. The term “comparative musicology” was supplanted by “ethnomusicology,” reflecting a new belief that cultural practices could only be understood in relation to the particular societies that gave rise to them: it was simply misleading to compare practices across different societies, the ethnomusicologists believed, and so the comparative musicologist was replaced by specialists in particular musical cultures. A similar reaction took place in theory and analysis: earlier style-analytical approaches (largely modeled on turn-of-the-century art history) gave way to a new emphasis on the particular structural patterns of individual musical works. Perversely, this meant that the possibility of computational approaches to the study of music arose just as the idea of comparing large bodies of musical data—the kind of work to which computers are ideally suited—became intellectually unfashionable. As a result, computational methods have up to now played a more or less marginal role in the development of the discipline. In this chapter I suggest that recent developments in computational musicology present a significant opportunity for disciplinary renewal: in the terms introduced in chapter 1, there is potential for musicology to be pursued as a more datarich discipline than has generally been the case up to now, and this in turn entails a re-evaluation of the comparative method. Central to any computational approach, however, are the means by which data are represented for analysis, and so I begin with some examples of “objective” data representations before introducing the issue of comparison. (The examples I discuss are graphic, but the same points could have been made in terms of numerical representations.) This is followed by an extended case study of an important current software package for musicological research, the Humdrum Toolkit, and the chapter concludes with a brief consideration of the prospects for computational methods in musicology.

103

104

(a)

(b)

(c)

(d)

105

Figure 6.1. Graphic analyses of Ligeti, Lux aeterna, section 2 (from Clendenning 1995: 948–49).

106

EMPIRICAL MUSICOLOGY

Issues of Representation A picture, as everyone knows, is worth a thousand words. Each of the graphs in Figure 6.1 (from Clendenning 1995: 248–249) represents a different aspect of the same section from Ligeti’s Lux aeterna: (a) charts pitch against time, giving an immediate impression of the music’s registral profile, (b) shows how many different pitches (not pitch classes) are present at any given point, and (c) highlights voice entries, with each separate voice (four each of soprano, alto, tenor, bass) having a separate horizontal line which is shaded whenever the voice enters, while (d) shows essentially the same information as (c), only in terms of the total number of entries at any given point. These graphs are “objective” in the sense that, once you have agreed how a graph is to be laid out, everyone should end up with the same result; the information is all there in the score, so that the analysis becomes, in effect, a matter of reformatting. But the decisions about how to lay them out are not necessarily self-evident. For example, in (a) to (c) the minimum values on the time axis are quavers, whereas in (d) they are whole bars; other values would have been possible, and the decision to use these particular values represents a judgment by the analyst that they will be the most informative in this context. There is also a degree of analytical judgment in the separate identification of the canons in (a) and (b). That said, however, it is obvious that there is scope for automation in graphs like this—and not just in drawing them up (the graphs in Figure 6.1 were generated by a computer drawing package) but in the processing of the data. The graphs in Figure 6.2 (from Brinkman and Mesiti 1991: 5, 6, 13, 17) were automatically generated from a machine representation of the score.1 Again, each graph represents a different way of extracting and viewing the information in a score, this time bars 1 to 26 of the first movement from Bartók’s String Quartet no. 4: (a) corresponds to Figure 6.1 (a), (b) shows each instrumental part separately with the vertical lines providing an impression of the linear movement from one pitch to another, and (c) charts the first appearance of each pitch, while (d) represents the dynamic level of each instrument (based on whether it is playing and on notated dynamic value) and of the whole (in effect a summation of the dynamic graphs of each instrument). But what are we to make of such representations? What do they enable us to see that we can’t hear, and in any case, what is the point of seeing music at all? It is a commonplace to describe Ligeti’s texture-oriented works as “visual,” and Brinkman and Mesiti (who in their article offer a case study of Webern’s Symphonie Op. 21) emphasize the spatial symmetry of Webern’s serial structures, commenting that “the graphs allow us to view all parts together in their spatial environment. [This] is especially helpful since the pitch symmetry that is essential to the musical structure of this work is somewhat obscured by the notated score” (Brinkman and Mesiti 1991: 21). The question then is the extent to which this problem of obscuring applies to other repertories, in other words how valuable this kind of visualization is for music in general. Yet this question is in some ways misguided: there is no such thing as a good or bad representation of music per se, there are good and bad — or at least better and worse —representations for particular purposes. To say that visual representations are more appropriate for twentieth-century than for other repertories, or for

Computational and Comparative Musicology

answering certain kinds of questions about the music than others, is simply to define their usefulness, not to criticize them. Nevertheless Brinkman and Mesiti produced some interesting results for music that one wouldn’t normally think of as particularly “visual,” including some strikingly different representations of music by Bach and Mozart as well as by Bartók, Schoenberg, Wolpe, and Berio. At the very least, such comparisons serve to underline the variety of textures found in different composers or styles— and texture is at the same time one of the most important aspects of music in terms of how we experience it, and one of the hardest to say anything useful about. It is not hard to imagine how Brinkman and Mesiti’s software could be used within a variety of pedagogical contexts, enabling students to literally and instantaneously “see” different pieces of music in a variety of different ways—and the ability to switch easily and quickly from one representation of music to another is a crucial element of practical musicianship. That this has not happened, and that the potential of the approach has not been fully realized, reflects the cost of translating academic research projects into software products for the real world, and illustrates an all-too-familiar pattern in musicological software: a sustained burst of initial enthusiasm is followed by running out of money, resulting in software that is sometimes less than fully functional, often less than fully documented, rarely properly supported, and usually soon obsolete. But my principal answer to the question “what are we to make of such representations?” is a different one. Brinkman and Mesiti (1991: 7) emphasize the particular usefulness of their graphs “as a device for comparing different pieces”; Matt Hughes, whose quantitative analysis of Schubert’s Op. 94 no. 1 was discussed in chapter 1, similarly claimed of his method that it was “a tool for organizing data so that one may more discernibly view tendencies and interassociations,” adding that it was “best utilized when viewing groups of compositions or sections of a composition rather than a single work” (Hughes 1977: 145, 150). And there is a general point to be made here: the more objective an analytical approach is, the more its musicological value is likely to be realized through making comparisons between different pieces—and sometimes large numbers of them. For example, when you carry out a Schenkerian analysis, you are in a sense making a comparison between the piece in question and the norms of common-practice voice leading, as systematized in the Schenkerian background and middleground. But the value of the analysis consists primarily in the lengthy process of making it, deciding which notes go with which, which are more important than others, and so forth; the process is lengthy because it involves a vast number of interpretive judgments, requiring you to weigh up different factors in relation to one another. At the end of it, you have a knowledge of the music—you might call it an intimacy—that you did not have at the outset, and there is a sense in which the final graph is significant mainly as a record of this learning process. With any kind of computational approach, by contrast, all of this happens automatically, and in some cases almost instantaneously; the only output is the graphic or numerical representation of the music that results. And such representations rarely tell you anything very useful by themselves: they may well be vulnerable to the cheap but telling jibe that they either show you what you hear anyhow (in which case they are redundant), or else they don’t (in which case there are

107

108

EMPIRICAL MUSICOLOGY

(a)

(b)

(c)

Computational and Comparative Musicology (d)

Figure 6.2. Graphic analyses of Bartók, String Quartet No. 4, I, bars 1–26 (from Brinkman and Mesiti 1991: 5, 6, 13, 17).

irrelevant). It is when you compare representations of different pieces that useful information may emerge. The value of objective representations of music, in short, lies principally in the possibility of comparing them and so identifying significant features, and of using computational techniques to carry out such comparisons speedily and accurately. At this point, however, the issue of reduction (already broached in chapter 1, this volume) has to be confronted. This can conveniently be explored in relation to the analysis of folksongs, since a large of proportion of early work in computational analysis was based on such material: there were substantial existing collections of folksongs to work on, and the music was monophonic and scalar, hence could be coded very compactly — a vital consideration in the days when computer memory was a scarce and expensive commodity. But of course, if you code folksongs as a series of pitches, or scale degrees, then you are leaving out all the variables of into-

109

110

EMPIRICAL MUSICOLOGY

Figure 6.3 EsAC code for “Deutschland über alles” (Helmut Schaffrath, Essen Musical Data Package)

nation, ornamentation, timbre, and other aspects of singing style, not to mention the original contexts of use and cultural connotations of the songs. And that obviously limits what you can conclude from comparing them. At this point a concrete example may be helpful, and Figure 6.3 is taken from the Essen Musical Data Package, a series of databases of popular and traditional vocal repertories together with tools for encoding and analysis.2 Coordinated until his death in 1994 by Helmut Schaffrath, the Essen databases consisted by the following year of some 10,000 songs, mainly European (and largely German), but with some representation of other traditions, particularly Chinese; they are still being added to, though development has been transferred from Essen to Warsaw (Selfridge-Field 1995). Figure 6.3 shows “Deutschland über alles” in the custom code used by the package, EsAC (the Essen Associative Code), and several of the data fields are self-explanatory: BOEHME is the name of the database in which this song is located, CUT contains the title, REG is the region from which it comes, and FCT is the functional category of the song, while KEY contains not only its key (F) but also the numbering of the song in the database (B0001), the value of the smallest note value (16, a sixteenth or semi-quaver), and the time signature (4 / 4). The tune itself is contained in the field MEL, using a simple code based on the scale degree (with # for sharp and b for flat degrees), ⫹ and ⫺ to indicate shifts to a higher or lower register, respectively, and a rhythmic notation based on the shortest value (semi-quaver), with a dot prolonging the value by 50 percent, a single underline character ( _ ) by 100 percent, and a double underline character ( __ ) by 200 percent. That should suffice for the tune to be read; the only other information contained in the MEL field is // (for the end of the melody), the use of spaces to indicate bar lines (the melody has been notated across the bar lines), and the splitting up of the melody into separate lines, each of which represents a phrase (this represents a judgment on the part of the transcriber, and as such is a rather unusual feature of the EsAC code). What can be done using this information? The distributed version of the Essen

Computational and Comparative Musicology

database comes with a simple analytical utility (ANA), which takes a complete database as its input, and generates as its output an annotated version of the file in which up to 12 different kinds of analytical information are added for each song. These include straightforward statistical information such as the distribution of ascending and descending intervals (how many ascending minor seconds?), the distribution of durations (how many crotchets, quavers, semi-quavers?), and distribution of scale degrees (how often does the fifth scale degree appear in the lower register?); there are also some derived measures such as the percentage of ascending steps and leaps. They also include some less obvious information, including the pattern of phrase repetitions in pitch and duration, a coding of the contour of each phrase, the melodic spine (based on the notes at metrically stressed beats), and the final notes of each phrase (this is where the coding of phrases comes in).3 A limitation of the ANA software is that it provides data for each song separately, whereas the statistical information would be more useful if it were provided across the database as a whole, but there is a reason for this: the package was intended for use with a commercial relational database package, which would facilitate this kind of analysis. Needless to say, the original package is no longer obtainable.4 But what can be done using this information? Schaffrath himself published a number of studies based on his datasets, in which he used this kind of analytical information to support broad stylistic generalizations, for example as between different traditions: “German folk songs,” he concluded from one such study (1992: 107), “generally skip more often up than down; in the Chinese pool the opposite applies. . . . German songs use the interval of a fourth 51% more often than Chinese songs do. This might be explained by the preference for pentatonic scales.” But there is also a different way in which such information can be used, which is for purposes of searching, identification, and classification. A database of 10,000 folksongs is not unlike a bibliographic database, and to locate individual items or groups of related items you search for particular features, combining them so as to narrow the search down to a manageable number of hits. In some contexts all you need to search a bibliographic database is the author’s name; in other cases you need to combine an author name, two title keywords, and year to locate the article you need. In the same way, you might try to locate a particular song by searching for its first few notes or incipit; this is the way that traditional dictionaries of themes operate (for instance, Barlow and Morgenstern 1983), and it works well with “art” music, where there are fixed scores. But it is not such a good way to search for folksongs, which typically exist in a large number of variants, often involving variation or the interpolation of notes. That is why ANA includes a routine to derive the final notes of each phrase: as Schaffrath (1997: 349) explains, “although variants may contain different numbers of notes or melodic contours, they tend to retain underlying harmonic structures.” Or you could combine such a search with a particular melodic spine and / or pattern of phrase repetition. And whereas I have described the use of such features in terms of simply locating items, the same processes can be used to group together songs or repertories in ways that might reflect, for example, historical development, linguistic associations, or conditions of use. In such ways the automated analysis of musical style can become a means of generating or verifying musicological hypotheses. Or at least that is the aspiration. In practice, EsAC and ANA suffer from obvious limitations, ranging from the pragmatic (the output of ANA is only marginally more

111

112

EMPIRICAL MUSICOLOGY

user-friendly than machine code) to those of principle. EsAC can cope with only a narrow range of music (monophonic, restricted to three octaves, and so on), and is well adapted for only a narrow range of applications. ANA extracts a limited number of features, each of which represents a radical reduction of the music and is intended for a particular purpose (such as statistical generalization or pattern matching); moreover, the usefulness of these reductions depends on a number of quite specific assumptions, such as that the final-note patterns of phrases vary less than incipits. All this makes sense in terms of what the Essen package is designed for; it is, to adopt outdated software jargon, a “turnkey” solution to folksong analysis, or at least to a certain sort of folksong analysis. And it would be possible to imagine trying to develop the code, and the analytical software, in such a way as to be more flexible — to cope with music of any desired degree of complexity and variety, and to extract from it every feature that could be of any possible interest under any possible circumstances. That would be like the approach of business software developers in the 1980s, who aimed at prepackaged solutions in which every eventuality had been foreseen and the appropriate feature bolted on somewhere, if only you could find it. In musicology, as perhaps also in the office, this represents a basic misapprehension—in terms not only of the unlimited variety of musical materials, but also of the variety of purposes for which people will want to represent them; as Huron (1992: 11) puts it, “The types of goals to which music representations may serve are legion and unpredictable.” Complexity is not the solution. There are, after all, much more complicated codes than EsAC which fulfill certain purposes extremely well while being very poorly adapted for others. MIDI code has transformed entire sectors of the music industry, but because it tries to understand all music on the model of keyboard music its handling of microtonal inflection is inelegant, to say the least; it also doesn’t distinguish enharmonic equivalents, and has to be drastically reconfigured if useful analytical information is to be extracted from it — even something as simple as locating a C major triad.5 Again, DARMS6 is an extremely powerful code designed for purposes of printing and publication (it is the basis of all A-R Edition scores), and accordingly it understands music as a kind of graphics; it doesn’t think of an e1 but of a notehead on the bottom line of a stave that has an F clef on it. As a result, it too needs drastic reconfiguration if analytically useful information is to be extracted from it (Brinkman 1986). And though the problems of multiple codes are alleviated by the existence of interchange utilities (e.g., to turn DARMS into MIDI), there are limitations to what is possible: since MIDI does not distinguish between C  and D
, the information is simply not there to translate into DARMS, which does make the distinction. As with business software, the solution to this problem of multiple requirements is not the creation of ever more complex and unwieldy integrated solutions, but a modular approach involving an unlimited number of individual software tools designed to serve individual purposes; that way, when you want to do something new, you simply design a new tool to do it. Modular approaches like this, however, need some kind of framework within which to work— in essence, a set of rules for the representation of data in machine-readable files, and for the transfer of information between different software tools. Humdrum is an example of just such a framework.

Computational and Comparative Musicology

A Case Study: The Humdrum Toolkit Developed by David Huron in the early 1990s, Humdrum7 consists of two distinct elements: in the first place a syntax—the set of rules I referred to — and in the second, a “toolkit” consisting at present of something over 70 separate software routines for manipulating the data in different ways for different purposes. It runs in a UNIX environment,8 and its modular approach is much more readily understood by those who have experience of UNIX than those who do not. Like Humdrum, UNIX can be thought of as a set of rules plus a set of tools: individual UNIX commands (which have names like grep, sort, and cat) do simple things like searching files for a particular string and extracting, deleting, or altering it, or rearranging data within or between files. The power and flexibility of the system comes not from these individual tools but from the possibility of “pipelining” them so that the output of one tool becomes the input of the next, forming chains of commands of unlimited length. The Humdrum toolkit is simply a set of additional UNIX tools designed specifically for manipulating music-related data. It is this open and modular philosophy, which always allows for an additional tool to be added when required, that explains why — as Huron (1997: 375) puts it—“Humdrum is especially helpful in music research environments where new and unforeseen goals arise.” Unfortunately, as we shall see, it also means that the user has to have a detailed understanding of the way the data are encoded and manipulated. Humdrum, in short, has a steep learning curve. All this may sound very abstract, so Figure 6.4 shows the first two phrases of “Deutschland über alles” in kern code, the normal Humdrum format for representing notated music (only two phrases because when printed out — though not in terms of internal storage—kern is much less compact than EsAC).9 This is a direct translation into kern of Figure 6.3, so the same information may be found in the data fields preceded by !!! (the exclamation marks identify them as comments); only the information concerning Schaffrath, the copyright statement, and the !!!ARL line (which gives latitudinal and longitudinal coordinates) are new. As for the other lines, **kern means that what follows is in kern (the double asterisk means that this label defines the data that follows), and the lines preceded by a single asterisk define the instrument category and instrument (in each case “vox” or voice), meter (4 / 4), key signature (B
, the “flat” being designated by “⫺”—the sign for sharp is “⫹”), and key (F).10 The tune itself begins on the following line, preceded by a curly bracket (this indicates the phrases, while the lines beginning with “⫽” are bar lines). “4.f” means that the first note is the F above middle C (if it were an octave lower it would be F, if an octave higher ff, if an octave higher than that fff, and so on); the “4” means that it is a quarter note (crotchet), and the “ . ”, unsurprisingly, that it is dotted. It should now be possible to read the melody quite straightforwardly. (The apparently simplistic reciprocal notation for durations is unexpectedly flexible, coping with most things short of Ferneyhough; triplet quavers, for instance, work out as 6, quintuplet semi-quavers as 20). That explains everything in Figure 6.4, except the “ *⫺ ”, which marks the end of the record. It should also explain everything in Figure 6.5, which shows the same phrases (with the comments omitted), as set in G major by Haydn in the second movement of his String Quartet Op. 76 no. 3: instead of the single column (or spine) of Figure 6.4, there are now four, one for each line of the

113

114

EMPIRICAL MUSICOLOGY

Figure 6.4 Kern code for first two phrases of “Deutschland über alles.”

music. It is helpful to think of this as being laid out like staff notation, only rotated by 90 degrees. Kern is not as easy to read as EsAC (which, apparently, was designed to be sight read). And it would seem to be better adapted for linear textures than, say keyboard music—though in fact there is no limitation in this respect, partly because you can put more than one note on a single line within any one spine, and also because you can collapse spines into one another or create new ones at any point. It is also more flexible than it looks. In the first place, you don’t have to specify everything in Figures 6.4 or 6.5; there is no need to encode phrases, or bar lines, or durations, or pitches (in fact the only mandatory lines are “**kern” and “*⫺”). In the second place, there is a large number of further codes and interpretations built into kern, al-

Computational and Comparative Musicology

Figure 6.5 Kern code for Haydn, String Quartet Op. 76, no. 3, II, bars 1–4

lowing you for instance to include symbols for articulation or ornamentation, clefs, stem direction, or the number of staff lines, if any (and whether they should be colored or dotted). But more than that, **kern spines can be mixed with others, including, for example, such predefined representations as **text (for lyrics) or **IPA (lyrics, but now notated using the International Phonetic Alphabet), or new representations defined by the user: you might, for example, create spines labeled **bowing, to show how a given pattern is (or might be) bowed, or **timing, to encode the rhythmic nuances in a particular performance, or **heartbeat, to record listeners’ physiological responses to hearing the music. Then again, you might not use **kern at all, but an alternative representation for pitch; predefined representations include **pitch (using the International Standards Organization format), **deg (scale degrees, as in EsAC), **solfg (French fixed-do solfège), **freq (frequency

115

116

EMPIRICAL MUSICOLOGY

expressed in cycles per second), or the self-explanatory **midi; any of these representations can be automatically translated to any other, provided of course that the relevant information is there to be translated (recall the problem with MIDI and enharmonic equivalents). There is also a particularly well-developed **fret representation, which turns pitch notation into tablature notation based on any specified number of strings and tuning, or of course you might always invent a new representation for some particular repertory or purpose. And what does all this enable you to do? For a start, you can replicate the functions of ANA: to extract the melodic spine of Figure 6.4, for instance, you locate the downbeat of each bar (search for lines beginning “⫽” and then skip to the next line), skip over the duration figures, and extract the pitch symbol(s) to a new file. Or you could just as easily search an entire “BOEHME” database for melodic motions from ^4 to ^5 and establish how frequently they occur by comparison with the reverse motion (you would translate the database into **deg records in order to do this, and search for adjacent lines containing 4 and 5, skipping over irrelevant symbols like durations, and finally counting the number found). Or you could search all of Bach’s chorales (the Riemenschneider collection is available in kern code) in order to establish how often the melodic leading-notes are approached from beneath, as against from above: to do this, you extract the spines containing the melodies, turn them into **deg records, and count the occurrences of “^7” as against “v7.”11 (This particular task is easier than it sounds, because **deg automatically codes the direction in which a given scale degree was approached, though not the size of the interval involved). Then again, shifting to a different level of complexity, you could analyze the relationship between a particular succession of vowels in song texts and the melodic or rhythmic contexts within which they occur. You could establish how far particular harmonic formations are correlated with specific metrical locations (and analyze the results by composer, period, or geographical origin). Or you could classify different melodic, harmonic, or structural contexts in Chopin’s complete mazurkas, and use this as the basis for analyzing timing information extracted from a large number of recordings. But the best way to see what Humdrum is capable of is to examine published studies that have used it, of which the majority (though by no means all) are by Huron and his coworkers; I shall briefly describe five. An article on the melodic arch in folksongs (Huron 1996) is based on the Essen collection and provides a direct comparison with Schaffrath’s own work. Its purpose is to test systematically the frequent, but inadequately supported, claim that “arch” contours are prevalent throughout Western folksongs, and Huron points out the dangers of “unintentional bias” (1996: 3) when such generalizations are supported by a small number of possibly handpicked examples: they can be firmly established only through the analysis of data sets large enough to allow the drawing of statistically significant conclusions. He puts forward two basic ways in which folksongs might be said to exhibit archshaped contours—within each phrase, and overall — and systematically analyzes the Essen collection for each. For the first test, individual phrases throughout the collection were sorted according to the number of notes in them, and the average contour for each phrase-length computed; for the second, the average pitch of each phrase of each song was computed, and the resultant contour classified. The overall

Computational and Comparative Musicology

conclusion was that there is a strong tendency toward arch-shaped contours on both measures (though a third approach, based on the nesting of arches within one another, was not supported). A project like this might be described as regulative: it takes an existing musicological claim and tests it rigorously against a large body of relevant data, adopting criteria of significance and certainty that may be unfamiliar in musicology but are normal in data-rich disciplines. Two further studies illustrate the range of musicological claims that can be tested in this manner. A study carried out in collaboration with Paul von Hippel (Hippel and Huron 2000) focused on the “gap-fill” model of melody, according to which listeners expect melodic leaps to be followed by a change of direction; this is an important element of Narmour’s “implication-realization” theory, which seeks to explain how listeners experience music on the basis of whether or not implications set up by the music are realized in what follows. What is at issue is not whether or not melodic leaps are usually followed by a change of direction, which is undoubtedly the case; it is whether or not this happens (as Krumhansl has concluded on the basis of experimental studies)12 because of listener expectations. The alternative is that it is a trivial consequence of the limited registral range of vocal music, a possibility which von Hippel and Huron (2000: 63) explain by comparison with a simplified melody that is confined to a range of three adjacent pitches — for example, A, B, and C. In such a melody, the only skip available is between A and C. Upward, this skip must land at the top of the range; downward, it must land at the bottom. After a skip, therefore, there is no way for a melody to continue moving in the same direction. On the contrary, two of the three available pitches can be reached only by a reversal. Although most melodies have a wider range, they are subject to the same basic argument. If this second possibility is the case, then gap-fill patterns will be no more common in real folksongs than in randomly generated melodies with otherwise comparable characteristics—and, to cut a long story short, this is exactly what von Hippel and Huron found. If their argument (which I have grossly simplified) is accepted, then it provides a striking instance of how studies based on large data sets can raise questions about conclusions derived even from experiments as meticulously controlled as Krumhansl’s.13 A third study (Huron 2001) makes a similar kind of point in relation to “art” music. It reassesses Allen Forte’s (1983) set-theoretical analysis of the first movement of Brahms’s String Quartet Op. 51, no. 1—an application to high Romantic music of an approach originally developed for twentieth-century music, bringing with it an appearance of objectivity and rigor in stark contrast to conventional motivic descriptions of such music. (In other words, Forte’s article is a late expression of the postwar culture of objectivity described in chapter 1, this volume.) Whereas a “motive” is simply a characteristic melodic figure that recurs prominently throughout a piece, Forte’s “alpha” ic (interval class) pattern can be defined much more precisely: it is a particular intervallic configuration that may occur in any of the standard transforms of set theory (prime, inversion, retrograde, retrograde inversion), in any rhythmic configuration, and at any metrical point. So Huron proceeds to search systematically for Forte’s “alpha” pattern, not only in the first movement of Op. 51,

117

118

EMPIRICAL MUSICOLOGY

no.1 but also in the first movements of Brahms’s other quartets — a comparable repertory in which the “alpha” pattern, if it is really characteristic of Op. 51, no. 1, should be significantly less common. First, he demonstrates that, as defined by Forte, the “alpha” pattern is common in Op. 51, no. 1, but no more so than in the other quartets. However, if you restrict the search to instances of the pattern that begin on metrically strong beats and come at the beginning of slurs, then it is more prevalent in Op. 51, no. 1 than in the others. This is particularly the case if you look for the pattern only in its prime form, and even more so if you consider only those cases where the pattern coincides with a long-short-long rhythmic pattern. So the prevalence of Forte’s “alpha” pattern is confirmed—but only under precisely those conditions that are captured by the conventional idea of a motive! The conclusion must be that the appearance of rigor in Forte’s analysis as against traditional analytical descriptions is just that, an appearance. Important and indeed salutary as this regulative function may be, it would be wrong to give the impression that computational analysis is good only for testing the validity of existing theoretical or analytical claims. It can also be used to undertake work that could hardly be achieved in any other way. A spectacular example is the correlation of musical features with geography, a project once again based on the Essen database (Aarden and Huron 2001). Recall that the Essen records include information concerning the geographical origins of the songs contained in them — sometimes down to the level of an individual town or village—and that the kern version of “Deutschland über alles” added longitudinal and latitudinal coordinates in the !!!ARL field (a new field defined specifically for Aarden and Huron’s project). Standard Humdrum commands can be used to extract records from the database and output their coordinates; these are then used as input to a mapping program. By way of example, Figure 6.6 shows the distribution of major- and minor-mode songs within the Essen database; these maps were generated using the GEO-Music site (http://www.music-cog.ohio-state.edu/cgi-bin/Mapping/map.pl, not accessible at publication time), established as part of the project. It is hard to know what conclusions might be drawn from this comparison (other than that the major mode is considerably more widespread than the minor); the southward skewing of the majormode data as against the minor simply reflects the three occurrences in Italy — hardly a basis for robust generalization—while the concentration in Germany and central Europe evident in both cases reflects the bias of the Essen database. A fully fledged musicological research project would no doubt involve dealing with more features at a greater level of detail. Nevertheless this example does indicate the potential for computational methods to draw together quite diverse kinds of information; it also illustrates the value of graphic presentation of the complex data generated by work in comparative musicology. A final study (Huron and Berec, forthcoming) is perhaps even more suggestive in terms of possible musicological applications. Like Hofstetter’s attempt (mentioned in chapter 1) to turn a claim about the “spirit of nationalism” into an empirically testable proposition, Huron’s and Berec’s starting point is a woolly, common-language concept: idiomaticism. How might you set about defining what is idiomatic on, say, the trumpet in such a way that a computer could make evaluations of just how idiomatic a particular piece of music is? In the first place, Huron and Berec say, idiomaticism is not the same as difficulty: a piece of trumpet music can be difficult but idio-

Figure 6.6 Distribution of major- and minor-mode songs (generated using the Geo-Music site) 119

120

EMPIRICAL MUSICOLOGY

matic, or easy and unidiomatic. It is rather a matter of achieving the desired musical effect in the easiest possible manner. For example, a piece will be idiomatic if it is significantly easier to play as written than when transposed up or down by say a semitone; this is “transposition idiomaticism.” So Huron and Berec evaluated the difficulty of a number of pieces when played in all transpositions from an octave below the original to an octave above, and in several cases the original proved to be significantly easier to play than most or all of the transposed versions. (They also ran a parallel test based on how much more difficult the music was when played at a different tempo from the notated one, with similar results.) But there was also a further finding: the pieces had been selected so as to include examples by both trumpet virtuosi and nontrumpeters, and the conclusion in respect of both transposition and tempo idiomaticism was the same—that expert trumpeters compose more idiomatically than composers who cannot play the instrument. That conclusion might be considered predictable, not to say obvious (just as in the case of Hofstetter’s discovery that there are differences between national styles). And it would have been of limited interest if Huron and Berec had simply asked trumpeters to play the pieces at different transpositions and at different tempos, and to say how difficult they were, an approach that would not have required Humdrum, of course. But instead—and this is the real point of their study — they attempted something much more challenging: what they refer to as “the design and implementation of a computer model of a trumpet performer.” Their model takes a kern file as its input, and evaluates the difficulty of performance along a variety of different dimensions: in terms of the valve transitions between successive notes; in terms of register, particularly in the case of sustained notes; in terms of dynamics; and in terms of tonguing at different speeds. (The information on which they based the model included performer assessments in the case of valve transitions, tests of performance in the case of tonguing, and physiological data concerning lung capacity and diaphragm support in the case of sustained notes.) The model was tested using a series of trumpet studies set for different grades by the Royal Conservatory of Music of Toronto; there was a generally fairly high correlation between its estimates of their difficulty and the Royal Conservatory’s. But of course the main test was the evaluation of idiomaticism, and the fact that the model come up with what everybody knows confirms the extent to which it actually works. That is the real conclusion to be drawn from the study. Do musicologists really need a computer model of a trumpet performer? One answer to this is that the approach could be readily generalized to other instruments: to the recorder (where finger transitions would be of particular importance, owing to the complexity of cross-fingerings), to the piano (the model might propose the best fingering, which could be tested against those specified by editors or used by performers), or to the guitar (in effect a modeling of the dance of the fingers on the fretboard). But the more substantial answer lies behind this. One of the chronic problems with analyzing music is the tendency toward abstraction: the more sophisticated our analytical models, the more divorced the object of analysis seems to become from the experience of music, and especially from the sense of physical engagement in which much if not all music has its source. It is conventional to deplore this, but to do nothing about it. In this context a computer model of the electric gui-

Computational and Comparative Musicology

tarist’s left hand could become, perhaps paradoxically, a means of thinking the body back into musical analysis: you could, to give just one example, chart the interaction of the dancing fingers with the ear—of the imperatives of the fretboard with those of “purely musical” implication and realization—in rock guitar improvisation. In this way computational musicology holds out the promise not only of providing more robust answers to questions that have already been asked, but also of making it possible to ask new questions.

Conclusion: Brave New World? Promises, promises . . . and it must be admitted that computational software has a hardly better record of delivering on its promises than dot com companies. Let’s stick with Humdrum as the most likely current candidate for a powerful, general-purpose computational aid for musicology, and explore some of the practicalities of doing musicology with it. Q. Is it easy to get hold of? A. Yes, at the time of writing you can download it for free, or you can buy it for the cost of the materials from the Center for Computer Assisted Research in the Humanities at Stanford.14 Q. Will it actually run on my computer? Don’t I have to be using a UNIX machine? A. You can run it on a PC or a Mac, but you will need to install a UNIX toolkit. You can probably get that for free, too.15 Q. But will I be able to find the music I want to work on in kern code? A. That depends what you want to work on. The Essen package is available in kern, and so is some of the Densmore collection of Native American music, so if you are interested in folk music you can start work right away. As for Western “art” music, there is quite a lot out there: not only Bach’s chorales but also most of his cantatas, as well as the Well-Tempered Clavier; substantial amounts of Corelli, Vivaldi, Handel, and Telemann; chamber and orchestral music by Haydn and Mozart; and most of the Beethoven symphonies.16 The hope, of course, is that as more musicologists use Humdrum, so more music will be encoded, encouraging more musicologists to use Humdrum—and in this way you get a virtuous circle. (The same applies to the development of the tools themselves.) Q. And if I can’t find what I want? A. If it exists in some other code you may be able to translate it into kern. Translation routines exist for the MuseData code used by the Center for Computer Assisted Research in the Humanities (CCARH), the Plaine and Easie Code used for RISM (Répertoire international des sources musicales), and MUSTRAN, as well as the code used by Leland Smith’s music notation program SCORE, and the “Enigma” code used by Finale. Not all of these translators are readily available or unproblematic, but if you still can’t find what you want you can of course encode your own. Rather than doing it by typing in the code, you can use an encoding routine that

121

122

EMPIRICAL MUSICOLOGY

enables you to enter the music on a MIDI keyboard — rather like steptime entry in a sequencer package. There are also utilities for playing kern code (via MIDI), for displaying music notation on screen, or for outputting it as a PostScript file (though not all of these are available if you are running Humdrum on a PC or a Mac). You can turn kern files into Finale ones for printing, too. Unfortunately, none of this addresses the real difficulties of integrating Humdrum into everyday musicological life. The fact is that not everybody is happy with a UNIX command-line environment, or finds it easy to remember the different commands with their multiple (and often complex) options; people who use UNIX (or Humdrum) on a daily basis can operate it much more quickly than a graphic environment (like Windows, with its drop-down menus and mouse control), but if Humdrum is to become part of everyday musicological life then what matters is its accessibility to the occasional user. Michael Taylor (1996) has developed a graphical user interface (GUI) for Humdrum, though it is not publicly distributed at the time of writing,17 and this makes life considerably easier for the occasional user: there are drop-down menus listing the various Humdrum commands and dialog boxes where you set the associated options, and you can even select musical elements using ordinary language (“any bar line,” “at least one flat,” or “C major chord,” for instance).18 For nonspecialist users such an interface—if generally available and supported — would surely represent a substantial step in the right direction: it not only looks familiar but also constantly reminds you of the commands and options that are available, without significantly compromising Humdrum’s functionality and flexibility. But such an approach can only go so far. This is because, to use Humdrum at all, you need quite a detailed knowledge of its representations (kern and the rest); a great deal of Humdrum usage has to do with things like extracting the spine you are interested in, merging spines, or stripping out information that is not wanted for any particular operation—and even ordinary-language definitions cannot protect you from the need to understand what is going on in the code. For this reason Andreas Kornstädt (1996) advocates a different approach: instead of “a Humdrum ‘command center’ with lots of buttons and gadgets with names like yank and MIDI|smf,” as he puts it (1996: 119), he has developed an analytical environment into which different GUI modules can be slotted for different applications. Each module can be thought of as a “browser” that let users view just those musical elements that they are interested in, and no others; for instance, Kornstädt has created such a system for leitmotivic analysis, in which users can define leitmotifs and locate their occurrence in a score on the basis of on-screen notation and intuitive commands. This, again, is not publicly available, but another of Kornstädt’s customized applications of Humdrum is: an on-line thematic dictionary, called Themefinder.19 You type in some pitches from the tune you want to find (or alternatively you can type in the scale degrees, or the intervals between the notes, or even just its contour), and the software finds all the matches in its database and displays them on screen in musical notation. There are no Humdrum commands; you don’t even need to know it is Humdrum that is doing the work—and what makes this possible is that Themefinder is designed to do one thing and one thing only. It is hard at the present time

Computational and Comparative Musicology

to tell how far Kornstädt’s approach—the development of specialized systems for different analytical purposes—will prove to be compatible with the flexibility and unpredictability of real-world musicological research, or whether it represents a particularly sophisticated way of falling into the same traps as 1980s office software. There is also the question of how the substantial costs of developing and updating a comprehensive system of this kind are to be found. Maybe, in the coming years, the combination of user-friendly interfaces for Humdrum and a developing body of musicological work using it will provide the necessary momentum for it to become more generally used, so that in time it will become just one of those skills that musicologists have to acquire (like using notation software, for example). Or maybe that is just not going to happen; in that case, computational software like Humdrum might still become part of the wider musicological scene, but as a specialist professional service, rather than a skill you acquire for yourself (rather like note processing was a decade or so ago). In other words, if you had a project in which computational approaches could play an important part—in generating or testing hypotheses on the basis of a large body of data, for instance —then you would call in the services of a specialist computational musicologist, who would arrange for any necessary encoding of data, carry out the analysis, and provide advice on the interpretation of the results. What the ordinary musicologist would need to know, then, would be not how to extract or collapse Humdrum spines, but how to formulate a research question in such a way that advantage can be taken of computational methods to answer it more securely than is possible using the methods traditional in data-poor fields. Whether anyone could make a decent living as a specialist computational musicologist is another matter. Using computers for musicology is like using computers for anything else: you need to have reasonable expectations about what they can do. However open and flexible its design, any musicological software is a tool, and like any tool it is good for some things and not for others. The articles by Huron and his coworkers that I have described in this chapter might be characterized as finding musicological uses to which computational tools can be put; in essence, they illustrate the software’s potential, and they do so in a manner that owes more to social-scientific discourse than to traditional humanities writing, with their explicit hypotheses, control groups, quantification of significance, and formulation in what Huron (1999b) has termed the “boilerplate language” of scientific inquiry. Perhaps that is only to underline the distinction between “cognitive” or “systematic” musicology, as such work is sometimes called, and the close attention to context and what might be termed epistemological pluralism of the traditional discipline. But what I am suggesting is that musicology in the broadest sense can take advantage of computational methods and transform itself into a data-rich discipline, without giving up on its humanist values. Understood that way, scientific discourse is unlikely to offer a general model for the musicology of the future: it is up to musicologists to develop such a model. And that does not mean trying to find musicological uses to which computational tools can be put, but the reverse: discovering the tools that are most appropriate to a particular musicological purpose, and taking full advantage of them. At present we hardly have such a thing as a model of how sophisticated computational approaches might be integrated within a sustained, musicologically driven project. Yet the tools are ready and waiting.20

123

124

EMPIRICAL MUSICOLOGY

Notes 1. The routines used to create these graphs, which took advantage of the NeXT computer’s Postscript display, have not to date been disseminated, but the principles underlying them were described in chapter 20 of Brinkman 1990. 2. Helmut Schaffrath, Essen Musical Data Package (Menlo Park, Calif.: Center for Computer Assisted Research in the Humanities [CCARH], 1995) [data and analytical software on four floppy discs for MS-DOS]); 6,225 folksongs were included in this release. Both data and software are now available at www.esac-data.org. (All URLs were correct at press time, but Web resources change rapidly. Search engines may provide a better means of access to the resources listed in this chapter.) 3. A full listing of the analytical output for “Deutschland über alles,” reformatted and annotated for clarity, may be found in Selfridge-Field 1995; see also Schaffrath 1992, 1997. 4. Since EsAC consists entirely of ASCII characters, it is possible to import the records into modern databases. However researchers now generally prefer to access the kern version of the databases and process them using Humdrum (see note 7 below). 5. Hence the need for translation programs such as POCO (see chapters 5 and 8, this volume). 6. Digital Alternate Representation of Musical Scores; for details see Selfridge-Field 1997, chapters 11– 15. 7. David Huron, The Humdrum Toolkit: Software for Music Researchers (Stanford, Calif.: Center for Computer Assisted Research in the Humanities, 1993 [three floppy discs and 16-page installation guide]). A wide range of information may be accessed from the Humdrum Toolkit home page (http://dactyl.som.ohio-state.edu/Humdrum; links to on-line resources are at http://dactyl.som.ohio-state.edu/Humdrum/resources.html). 8. UNIX is an operating system, like MS-DOS or Windows, but includes a large number of general-purpose tools and thus fulfils many of the functions of a programming environment as well. Developed in the late 1960s, it remains the environment of choice for many professional software developers and researchers. 9. For a summary of kern see Huron 1997; for detailed descriptions see Huron 1995, 1999a. 10. Lines beginning with asterisks are known as “interpretation records,” those with two asterisks being inclusive interpretations (something is either in kern or it isn’t), and those with one asterisk being tandem interpretations (of which you can have any number). 11. In a review of the Humdrum Toolkit, Jonathan Wild (1996) includes what is essentially a tutorial on this precise task; assuming that the chorales have been concatenated into one file, the entire analysis consists of seven commands pipelined to another — in other words, it can be executed with a single command line. Further analytical applications of Humdrum are discussed in Huron 2002, which appeared after this chapter had gone into production. 12. See this volume, pp. 7–8. 13. A follow-up study (Hippel 2000) reanalyzed the experimental data on the basis of which Burton Rosner and Leonard Meyer had claimed that gap-fill patterns reflect listener expectations; von Hippel concludes that that the apparent effects of gap-fill expectations were an artifact of experimental design (specifically, effects of training). Of course listeners may have such expectations, but von Hippel’s and Huron’s point is that these are the result of melodic patterns and not the other way around.

Computational and Comparative Musicology 14. See the Humdrum download page at http://dactyl.som.ohio-state.edu/ HumdrumDownload/downloading.html. Additional tools developed by Craig Sapp can be downloaded from http:/ / www.ccarh.org / software /humdrum/museinfo. 15. UWIN (free to educational/research users) is accessible through the Humdrum download page (see n. 14). Commercial alternatives are available from Morton Kern Systems (MKS Toolkit, like UWIN for Windows) or from Tenon Intersystems (for Mac). There is a further alternative: a number of Humdrum commands can be run online at http:/ / musedata.stanford.edu / software / humdrum/online; you select the command you want, supply the input data (by pasting it into a window, uploading a file, or supplying its URL), set the options, and receive the output as plain text or in HTML. Twenty of the 70 or more Humdrum commands are currently available in this way, with an emphasis on data translation and conversion. (Not accessible at publication time.) 16. There are two principal sources for such material (from both of which they may be obtained free): the KernScores site at http:/ / kern.humdrum.net, and CCARH’s MuseData site (http:/ / www.musedata.org). 17. “Humdrum Toolkit GUI” currently exists only in a 16-bit version, but further development (and eventual distribution) is planned by the Sonic Arts Centre at Queen’s University, Belfast. 18. Only a few of these “regular expressions,” as they are known in UNIX jargon, are built into the interface, but you can add your own, along with new interpretations; this is where you would define **timing or **heartbeat. 19. Accessible at http:/ / www.themefinder.org; for a brief description see Kornstädt 1998. 20. My grateful thanks to David Huron, who has influenced this chapter in many ways, not least through making it possible for me to work with him in 2000 as a visiting scholar at the Cognitive Musicology Laboratory, Ohio State University, Columbus.

References Aarden, B., and Huron, D. (2001). “Mapping European folksong: Geographical localization of musical features,” in Walter B. Hewlett and Eleanor Selfridge Field (eds.), The Virnal Score: Representation, Retrieval, Restoration (Computing in Musicology 12). Cambridge, Mass.: MIT Press, 169 – 183. Barlow, H., and Morgenstern, S. (1983). A Dictionary of Musical Themes, rev. ed. London: Faber. Brinkman, A. R. (1986). “Representing musical scores for computer analysis.” Journal of Music Theory 30: 225 – 275. Brinkman, A. R. (1990). Pascal Programming for Music Research. Chicago: Chicago University Press. Brinkman, A. R., and Mesiti, M. R. (1991). “Graphic modeling of musical structure.” Computers in Music Research 3: 1–42. Clendenning, J. P. (1995). “Structural factors in the microcanonic compositions of György Ligeti,” in E. W. Marvin and R. Hermann (eds.), Concert Music, Rock, and Jazz since 1945: Essays and Analytical Studies. Rochester, N.Y.: University of Rochester Press, 229–256. Forte, A. (1983). “Motivic design and structural level in the First Movement of Brahms’s String Quartet in C Minor.” Musical Quarterly 69: 471 – 502.

125

126

EMPIRICAL MUSICOLOGY

Hippel, P. von (2000). “Questioning a melodic archetype: Do listeners use gap–fill to classify melodies?” Music Perception 18: 139 – 153. Hippel, P. von, and Huron, D. (2000). “Why do skips precede melodic reversals? The effect of tessitura on melodic structure.” Music Perception 18: 59–85. Hughes, M. (1977). “[Schubert, Op. 94 No. 1:] a quantitative approach,” in M. Yeston (ed.), Readings in Schenker Analysis and Other Approaches. New Haven: Yale University Press, 144–164. Huron, D. (1992). “Design principles in computer-based music representation,” in A. Marsden and A. Pople (eds.), Computer Representations and Models in Music. London: Academic Press, 5 – 39. Huron, D. (1995). The Humdrum Toolkit: Reference Manual. Menlo Park, Calif.: Center for Computer Assisted Research in the Humanities [accessible online at http://dactyl .som.ohio-state.edu / Humdrum / commands.toc.html]. Huron, D. (1996). “The melodic arch in Western folksongs.” Computing in Musicology 10: 3–23. Huron, D. (1997). “Humdrum and Kern: Selective feature encoding,” in E. Selfridge-Field (ed.), Beyond MIDI: The Handbook of Musical Codes. Cambridge, Mass.: MIT Press, 375–401. Huron, D. (1999a), Music Research Using Humdrum: A User’s Guide. Stanford, Calif.: Center for Computer Assisted Research in the Humanities [accessible online at http:/ /dactyl.som.ohio-state.edu / Humdrum / guide.toc.html]. Huron, D. (1999b). “Methodology: On Finding Field-Appropriate Methodologies at the Intersection of the Humanities and the Social Sciences.” 1999 Ernest Bloch lectures, University of California at Berkeley, #3 [accessible online at http://dactyl.som .ohio-state.edu / Music220 / Bloch.lectures / 3.Methodology.html]. Huron, D. (2001). “What is a musical feature? Forte’s analysis of Brahms’s Opus 51, No. 1, Revisited.” Music Theory Online 7.4. Huron, D. (2002). “Music information processing using the Humdrum Toolkit: Concepts, examples, and lessons.” Computer Music Journal 26/2: 11–26. Huron, D., and J. Berec (forthcoming). “Characterizing idiomatic organization in music: A theory and case study.” Journal of New Music Research. Kornstädt, A. (1996). “SCORE-to-Humdrum: A graphical environment for musicological analysis.” Computing in Musicology 10: 105 – 122. Kornstädt, A. (1998). “Themefinder: A web-based melodic search tool.” in Walter B. Hewlett and Eleanor Selfridge-Field (eds.), Melodic Similarity: Concepts, Procedures, and Applications (Computing in Musicology, 11). Cambridge, Mass.: MIT Press, 231–236. Schaffrath, H. (1992). “The retrieval of monophonic melodies and their variants: Concepts and strategies for computer-aided analysis,” in A. Marsden and A. Pople (eds.), Computer Representations and Models in Music. London: Academic Press, 95–109. Schaffrath, H. (1997). “The Essen Associative Code,” in Eleanor Selfridge-Field (ed.), Beyond MIDI: The Handbook of Musical Codes. Cambridge, Mass.: MIT Press, 343–361. Selfridge-Field, E. (1995). Essen Musical Data Package (CCARH Technical Report No. 1), Menlo Park, Calif.: Center for Computer Assisted Research in the Humanities [distributed with the Essen Musical Data Package]. Selfridge-Field, E. (1997). Beyond MIDI: The Handbook of Musical Codes. Cambridge, Mass.: MIT Press. Taylor, W. M. (1966). “Humdrum Graphical User Interface.” M.A. dissertation., Music Technology program, Queen’s University of Belfast. Wild, J. (1996). “A review of the Humdrum Toolkit: UNIX tools for music research, created by David Huron.” Music Theory Online 2.7.

CHAPTER 7

Modeling Musical Structure Anthony Pople

More than 20 years ago, music analysis was famously described by Ian Bent as “that part of the study of music which takes as its starting-point the music itself, rather than external factors” (Bent 1980: 341). Indeed, analysis is generally motivated by a desire to encounter a piece of music more closely, to submit to it at length, and to be deeply engaged by it, in the hope of thereby understanding more fully how it makes its effect. It is perhaps not surprising, then, that if you were to take a look at the kinds of writing that have at one time or another been thought of as music analysis, the variety would be immense. To a large extent this is because of the personal element in analysis: a piece of analytical writing is almost always the work of one person, and is founded in that person’s own experience of an individual work. But even in this regard music analysis is significantly different from music criticism, or indeed literary criticism, because analysts generally try to play down the fact that their analyses are dependent on a personal viewpoint. Although music analysis certainly does have a critical dimension as one of its characteristic attributes (see Pople 1994), a writer of music analysis will in general try to present his or her observations as representing a musical experience that can be shared, and will address the reader as a kindred spirit eager to inquire about the piece in the same terms as the author has done. This balance between the personal point of view and the potential for capturing shared experience makes music analysis a domain that is not only by its very nature empirical, but also one on which formal empirical methods can be brought to bear. However, the fact that one can say this in relation to analysis as the term is understood today is very much a consequence of the prevailing close relationship between music analysis and music theory. As Nicholas Cook has pointed out, there is a broad historical distinction between music theory—which studies musical works in order to deduce “more general principles of musical structure”—and music analysis, in which the interest is focused on individual pieces of music (Cook 1987: 7). Since about 1970, without contradicting this distinction, a symbiotic relationship has arisen between theory and analysis: theory has developed by using analysis as a kind of test-bed, while professional analysts have by and large conscientiously used the language of contemporary theory to express their insights. This scenario is sufficiently close to the scientific model of hypothesis and experiment to have aroused antagonism from some humanities scholars (see, for example, Snarrenberg 1994: 50 – 55). But it has also allowed other scholars from disciplines that are more committed to formal methodol-

127

128

EMPIRICAL MUSICOLOGY

ogies, including cognitive psychologists and computer scientists, to see music analysis as a potentially fruitful domain of research. Some of what has been achieved is remarkable, but from a music analyst’s point of view the question is: can it engage the reader’s musical interest as deeply as can analytical writing based on expert human contemplation? It is important to bear this in mind if one is to weigh up the value of differing approaches in which the empirical quality so fundamental to analysis is treated more formally. I have written elsewhere that “when people read something called ‘music analysis’ they assume that they will be ‘told something about the piece’,” and that in this respect there are specifically two types of failure that can occur if the analysis fails to engage sufficiently with the personal, empirical, critical and didactic aspects of analysis: The kind of . . . writing that remains tangled up in its own high-level assumptions, and so deals with musical detail in an anecdotal fashion . . . certainly “tells us something”; but, to those who expect the strategic sweep of analysis to range constructively through to the personal / empirical, what it tells us may not seem to be “about the piece.” Conversely, work that is excessively bottom-up . . . may remain tangled in the empirical and fail to cross other than trivially into the critical/didactic domain; it is certainly “about the piece,” but may “tell us” little or nothing. (Pople 1994: 121) As Cook points out, good analysis does not merely reflect and describe experience, but also has the ability to make us hear the music differently (Cook 1987: 228 – 229). Together, these observations constitute a set of linked criteria that should be borne in mind as we examine in detail such issues as the place of rigorous technical language in music theory, the quasi-scientific approach to the relationship between theory and analysis, and the modeling of music analysis from interdisciplinary perspectives.

Formal Theory, Informal Analysis Take, for example, the development of a rigorous terminology for the description of serial music at Princeton University in the 1960s, under the influence of the mathematically trained composer and theorist Milton Babbitt. In an article on “TwelveTone Rhythmic Structure and the Electronic Medium” published in the inaugural issue of the journal Perspectives of New Music, Babbitt outlined a theory of serial rhythm using language that mathematicians would regard as informal, but which to classically trained musicians might seem mathematical: [As] a means of informally evaluating the temporal constraints imposed by the formation of a twelve-tone set, I shall assume on purely empirical grounds that there are eleven qualitatively significant temporal relationships which can hold between two musical (say, pitch) events. Let x and y designate these events, and let a left parenthesis signify the time point initiation of the event and a right parenthesis signify the time point termination . . .

Modeling Musical Structure

1. x) < (y. [that is, the termination of x precedes the initiation of y] 2. (x < (y; x) < y); but x)